From 46cffd863c8aa8fcc4fe0b6d607397f6c21c5520 Mon Sep 17 00:00:00 2001
From: Nelly Barret <nelly.barret@etu.univ-lyon1.fr>
Date: Mon, 20 Jul 2020 19:13:13 +0200
Subject: [PATCH] [AM] documentation is finished and generated
---
doc/Data.html | 968 +++++++++++++
doc/Dataset.html | 515 +++++++
doc/Method.html | 178 +++
doc/MethodCleaning.html | 705 +++++++++
doc/MethodPrediction.html | 282 ++++
doc/MethodSelection.html | 272 ++++
doc/__init__.html | 57 +
doc/cleaning.html | 275 ++++
doc/config.html | 153 ++
doc/main.html | 683 +++++++++
doc/model.html | 490 +++++++
doc/predict.html | 345 +++++
doc/selection.html | 406 ++++++
doc/utility_functions.html | 1264 +++++++++++++++++
predihood/classes/Dataset.py | 63 +-
predihood/classes/Method.py | 16 +-
predihood/classes/MethodCleaning.py | 55 +-
predihood/classes/MethodPrediction.py | 7 +-
predihood/classes/MethodSelection.py | 33 +-
predihood/cleaning.py | 18 +-
.../distribution-plots/distribution_geo.png | Bin 28979 -> 28987 bytes
predihood/main.py | 60 +-
predihood/model.py | 54 +-
predihood/predict.py | 71 +-
predihood/selection.py | 129 +-
predihood/utility_functions.py | 395 +++---
26 files changed, 7070 insertions(+), 424 deletions(-)
create mode 100644 doc/Data.html
create mode 100644 doc/Dataset.html
create mode 100644 doc/Method.html
create mode 100644 doc/MethodCleaning.html
create mode 100644 doc/MethodPrediction.html
create mode 100644 doc/MethodSelection.html
create mode 100644 doc/__init__.html
create mode 100644 doc/cleaning.html
create mode 100644 doc/config.html
create mode 100644 doc/main.html
create mode 100644 doc/model.html
create mode 100644 doc/predict.html
create mode 100644 doc/selection.html
create mode 100644 doc/utility_functions.html
diff --git a/doc/Data.html b/doc/Data.html
new file mode 100644
index 00000000..86a0e887
--- /dev/null
+++ b/doc/Data.html
@@ -0,0 +1,968 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>Data API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>Data</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import sys
+
+import numpy as np
+import os
+import pandas as pd
+import warnings
+from area import area
+
+from predihood import model
+from predihood.cleaning import clean
+from predihood.config import FILE_CLEANED_DATA, FOLDER_DATASETS, ENVIRONMENT_VARIABLES, FILE_GROUPING, NEW_PREFIX, OLD_PREFIX
+from predihood.utility_functions import address_to_code, append_indicator, append_target
+
+log = logging.getLogger(__name__)
+warnings.filterwarnings("ignore", category=RuntimeWarning)
+
+
+class Data:
+ def __init__(self, normalization="density", filtering=True):
+ """
+ Constructor of the Data class. Initialize attributes.
+ Args:
+ normalization: A string to indicate the choice for normalization ("density" for density, "population" for population and None to do not normalize)
+ filtering: True or False to indicate if useless indicators will be removed or not
+ """
+ self.old_path = FILE_CLEANED_DATA
+ self.dataset_path = os.path.join(FOLDER_DATASETS, "data_"+str(normalization)+".csv")
+ self.filtered_path = os.path.join(FOLDER_DATASETS, "data_"+str(normalization)+"_filtered.csv")
+ self.data = None
+ self.indicators = None
+ self.normalization = normalization
+ self.filtering = filtering
+
+ # retrieve indicators
+ self.get_indicators()
+ log.debug("Starting with %d indicators", len(self.indicators))
+ # define indicators that are not relevant for the prediction and remove them
+ indicators_to_remove = ["IRIS", "REG", "DEP", "UU2010", "COM", "LIBCOM", "TRIRIS", "GRD_QUART", "LIBIRIS", "TYP_IRIS", "MODIF_IRIS", "LAB_IRIS", "LIB_IRIS", "LIB_COM", "CODGEO", "LIBGEO"]
+ for indicator in indicators_to_remove:
+ if indicator in self.indicators:
+ self.indicators.remove(indicator)
+ log.debug("Descriptive indicators: %d. It remains %d indicators", len(indicators_to_remove), len(self.indicators))
+
+ def get_indicators(self):
+ """
+ Get indicators from the dataset if it exists, else from the database.
+ """
+ if self.data is not None:
+ self.indicators = self.data.columns.tolist()
+ # self.data.columns gets all columns, so to get indicators, we remove "CODE" (since it is not relevant for prediction).
+ # We keep "AREA" and "DENSITY" since they are use as features in prediction
+ if "CODE" in self.indicators: self.indicators.remove("CODE")
+ for env in ENVIRONMENT_VARIABLES:
+ if env in self.indicators:
+ self.indicators.remove(env)
+ else:
+ self.indicators = model.get_indicators_list()
+
+ def init_all_in_one(self):
+ """
+ Create or read the dataset. It applies normalization and/or filtering if needed.
+ """
+ if not os.path.exists(self.dataset_path):
+ self.create()
+ self.apply_normalization()
+ if self.filtering: self.apply_filtering()
+ elif not os.path.exists(self.filtered_path):
+ if self.filtering: self.apply_filtering()
+ else:
+ self.read()
+
+ def read(self):
+ """
+ Read the dataset stored in the CSV file (and get indicators from this dataset).
+ """
+ if self.filtering:
+ self.data = pd.read_csv(self.filtered_path)
+ else:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()
+
+ def create(self):
+ """
+ Construct the dataset from HomeInLove data, i.e. ungroup addresses, retrieve information about IRIS and construct a CSV file.
+ """
+ log.info("... construction of dataset is in progress ...")
+
+ # 1. read data to transform
+ if not os.path.exists(self.old_path): clean()
+ data = pd.read_csv(self.old_path)
+
+ # 2. define some variables
+ raw_data = []
+ append_col = True
+ null_iris = 0 # count IRIS that can't be retrieve from its code
+ foreigners = 0 # count addresses that are not located in France
+ problems = 0 # count IRIS that encounter problems and that will be not added to the dataset
+ cols = []
+ columns = ["address", "country"]
+ columns.extend(ENVIRONMENT_VARIABLES)
+ cols_departure = [OLD_PREFIX+env for env in columns]
+ cols_arrival = [NEW_PREFIX+env for env in columns]
+
+ # 3. get data for departure and arrival IRIS
+ departures = data[cols_departure]
+ arrivals = data[cols_arrival]
+ # remove "old_address", "old_country", "new_address", "new_country"
+ cols_departure.pop(0)
+ cols_departure.pop(0)
+ cols_arrival.pop(0)
+ cols_arrival.pop(0)
+ cols.append("CODE")
+ cols.append("AREA")
+ cols.append("DENSITY")
+
+ # 4. for each departure and arrival IRIS (one line in cleanedData.csv), get its INSEE indicators and its EV.
+ # for each IRIS store its code, compute and store its area and its density, store all its INSEE indicators and its 6 EV.
+ for (index1, row1), (index2, row2) in zip(departures.iterrows(), arrivals.iterrows()):
+ dep_row, arr_row = [], []
+ # convert address to IRIS code and append it to data only if the address is in France
+ if row1[OLD_PREFIX + "country"] == "France":
+ dep_code = address_to_code(row1[OLD_PREFIX + "address"])
+ dep_row.append(dep_code)
+ dep_iris = model.get_iris_from_code(dep_code) # get IRIS information
+ coords_dep = model.get_coords_from_code(dep_code)
+ area_dep = area(coords_dep) / 1000000 if coords_dep is not None else None # convert area from m^2 to km^2
+ dep_row.append(area_dep) # add area as a feature
+ density = dep_iris["properties"]["raw_indicators"]["P14_POP"] / area_dep if area_dep is not None and area_dep > 0 else None
+ dep_row.append(density) # add density as a feature
+
+ if dep_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ dep_row, cols = append_indicator(raw_indicator, dep_iris, dep_row, append_col, cols)
+
+ # b. append EV
+ for target in cols_departure:
+ dep_row = append_target(row1, target, dep_row)
+ append_col = False
+ if len(dep_row) > 0: raw_data.append(dep_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row1[OLD_PREFIX+"country"] == "" or row1[OLD_PREFIX+"country"] is None or row1[OLD_PREFIX+"country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+ if row2[NEW_PREFIX + "country"] == "France":
+ arr_code = address_to_code(row2[NEW_PREFIX + "address"])
+ arr_row.append(arr_code)
+ arr_iris = model.get_iris_from_code(arr_code)
+ coords_arr = model.get_coords_from_code(arr_code)
+ area_arr = area(coords_arr) / 1000000 if coords_arr is not None else None # convert area from m^2 to km^2
+ arr_row.append(area_arr) # add area as a feature
+ density = arr_iris["properties"]["raw_indicators"]["P14_POP"] / area_arr if area_arr is not None and area_arr > 0 else None
+ arr_row.append(density) # add density as a feature
+
+ if arr_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ arr_row, cols = append_indicator(raw_indicator, arr_iris, arr_row, append_col, cols)
+
+ # b. append targets
+ for target in cols_arrival:
+ arr_row = append_target(row2, target, arr_row)
+ append_col = False
+ if len(arr_row) > 0: raw_data.append(arr_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row2[NEW_PREFIX + "country"] == "" or row2[NEW_PREFIX + "country"] is None or row2[NEW_PREFIX + "country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+
+ sys.stdout.write("\r%.2f%%" % ((index1 * 100) / len(departures))) # update progress percentage
+ sys.stdout.flush()
+ print()
+
+ cols.extend(ENVIRONMENT_VARIABLES)
+ log.info("%d addresses are not located in France.", foreigners)
+ log.info("%d null IRIS have been removed from the dataset.", null_iris)
+ log.info("%d IRIS have encountered problems.", problems)
+
+ # 5. convert IRIS data into a DataFrame, fill missing values and remove fully empty columns.
+ self.data = pd.DataFrame(raw_data, columns=cols)
+ self.data.sort_values("CODE", inplace=True)
+ self.fill_missing_values("median")
+ nan_columns = self.data.columns[self.data.isna().all()].tolist()
+ log.debug("There are %d NaN columns", len(nan_columns))
+ self.data.drop(nan_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in nan_columns] # remove names of NaN columns in the list of indicators
+
+ def fill_missing_values(self, method="median"):
+ """
+ Fill NaN values given the method.
+ Args:
+ method: A string corresponding tto the method for filling NaN values ("zero", "mean" or "median"). Default is median.
+ """
+ assert method in ["zero", "mean", "median"]
+ cols = ["CODE", "AREA", "DENSITY"]
+ cols.extend(ENVIRONMENT_VARIABLES)
+ for col in self.data.iteritems():
+ value = 0
+ col_name = col[0]
+ if col_name not in cols:
+ # fill missing INSEE indicators
+ if method == "zero":
+ value = 0
+ elif method == "mean":
+ value = self.data[col_name].mean()
+ elif method == "median":
+ value = self.data[col_name].median()
+ else:
+ value = np.mean(self.data[col_name])
+ elif col_name not in ["CODE", "AREA", "DENSITY"]:
+ # fill missing EV
+ if method == "zero":
+ env_values = {"building_type": "Housing estates",
+ "building_usage": "Other activities",
+ "landscape": "Urban",
+ "morphological_position": "Central",
+ "geographical_position": "Centre",
+ "social_class": "Lower"}
+ else: # method == "mean" or method == "median"
+ env_values = {"building_type": "Mixed",
+ "building_usage": "Other activities",
+ "landscape": "Green areas",
+ "morphological_position": "Urban",
+ "geographical_position": "Centre",
+ "social_class": "Middle"}
+ value = env_values[col_name]
+ # else: -> column is CODE or AREA or DENSITY -> do nothing
+ self.data[col_name].replace([np.nan], [value], inplace=True)
+
+ def filter_too_detailed_indicators(self):
+ """
+ Remove too detailed indicators from the dataset. The list of indicators is given by the file `regrouping.csv`.
+ """
+ regrouping = pd.read_csv(FILE_GROUPING, sep="\t")
+ status = regrouping.index[regrouping["STATUS"] == 1].tolist() # indicators that are an element of a subset
+ col_names = [regrouping.iloc[status]["INDICATOR"] for status in status]
+ # self.filtered_data = self.data[:] # copy by value, not by reference
+ counter1, counter2 = 0, 0
+ for column in col_names:
+ if column in self.data and column in self.indicators and column:
+ del self.data[column]
+ self.indicators.remove(column)
+ counter1 += 1
+ log.debug("%d indicators have been using regrouping.", counter1)
+
+ def apply_filtering(self):
+ """
+ Remove (filter) too detailed indicators and constant columns from the dataset.
+ """
+ if self.data is None:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()
+
+ # 1. remove indicators that have been defined as useless
+ self.filter_too_detailed_indicators()
+
+ # 2. remove constant columns, i.e. with a null variance
+ constant_columns = self.data.columns[self.data.nunique() == 1].tolist()
+ self.data.drop(constant_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in constant_columns]
+
+ self.data.to_csv(self.filtered_path, index=None) # index=None: avoid line numbering
+
+ def apply_normalization(self):
+ """
+ Normalize the dataset with the given method, i.e. None, "population" or "density". Default is "density".
+ """
+ assert self.normalization in [None, "population", "density"]
+ do_not_normalize = ["CODE", "AREA", "DENSITY", "P14_POP"]
+ do_not_normalize.extend(ENVIRONMENT_VARIABLES) # extend does not return a list
+ for index, row in self.data.iterrows():
+ for column in row.iteritems():
+ col_name = column[0]
+ if col_name not in do_not_normalize:
+ if self.normalization == "population":
+ self.data.at[index, col_name] = row[col_name] / row["P14_POP"]
+ elif self.normalization == "density":
+ density = row[self.data.columns.get_loc("DENSITY")]
+ self.data.at[index, col_name] = row[col_name] / density if density > 0 else 0
+ elif self.normalization is None:
+ self.data.at[index, col_name] = row[col_name]
+ self.data.to_csv(self.dataset_path, index=None) # index=None: avoid line numbering</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="Data.Data"><code class="flex name class">
+<span>class <span class="ident">Data</span></span>
+<span>(</span><span>normalization='density', filtering=True)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Constructor of the Data class. Initialize attributes.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>normalization</code></strong></dt>
+<dd>A string to indicate the choice for normalization ("density" for density, "population" for population and None to do not normalize)</dd>
+<dt><strong><code>filtering</code></strong></dt>
+<dd>True or False to indicate if useless indicators will be removed or not</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class Data:
+ def __init__(self, normalization="density", filtering=True):
+ """
+ Constructor of the Data class. Initialize attributes.
+ Args:
+ normalization: A string to indicate the choice for normalization ("density" for density, "population" for population and None to do not normalize)
+ filtering: True or False to indicate if useless indicators will be removed or not
+ """
+ self.old_path = FILE_CLEANED_DATA
+ self.dataset_path = os.path.join(FOLDER_DATASETS, "data_"+str(normalization)+".csv")
+ self.filtered_path = os.path.join(FOLDER_DATASETS, "data_"+str(normalization)+"_filtered.csv")
+ self.data = None
+ self.indicators = None
+ self.normalization = normalization
+ self.filtering = filtering
+
+ # retrieve indicators
+ self.get_indicators()
+ log.debug("Starting with %d indicators", len(self.indicators))
+ # define indicators that are not relevant for the prediction and remove them
+ indicators_to_remove = ["IRIS", "REG", "DEP", "UU2010", "COM", "LIBCOM", "TRIRIS", "GRD_QUART", "LIBIRIS", "TYP_IRIS", "MODIF_IRIS", "LAB_IRIS", "LIB_IRIS", "LIB_COM", "CODGEO", "LIBGEO"]
+ for indicator in indicators_to_remove:
+ if indicator in self.indicators:
+ self.indicators.remove(indicator)
+ log.debug("Descriptive indicators: %d. It remains %d indicators", len(indicators_to_remove), len(self.indicators))
+
+ def get_indicators(self):
+ """
+ Get indicators from the dataset if it exists, else from the database.
+ """
+ if self.data is not None:
+ self.indicators = self.data.columns.tolist()
+ # self.data.columns gets all columns, so to get indicators, we remove "CODE" (since it is not relevant for prediction).
+ # We keep "AREA" and "DENSITY" since they are use as features in prediction
+ if "CODE" in self.indicators: self.indicators.remove("CODE")
+ for env in ENVIRONMENT_VARIABLES:
+ if env in self.indicators:
+ self.indicators.remove(env)
+ else:
+ self.indicators = model.get_indicators_list()
+
+ def init_all_in_one(self):
+ """
+ Create or read the dataset. It applies normalization and/or filtering if needed.
+ """
+ if not os.path.exists(self.dataset_path):
+ self.create()
+ self.apply_normalization()
+ if self.filtering: self.apply_filtering()
+ elif not os.path.exists(self.filtered_path):
+ if self.filtering: self.apply_filtering()
+ else:
+ self.read()
+
+ def read(self):
+ """
+ Read the dataset stored in the CSV file (and get indicators from this dataset).
+ """
+ if self.filtering:
+ self.data = pd.read_csv(self.filtered_path)
+ else:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()
+
+ def create(self):
+ """
+ Construct the dataset from HomeInLove data, i.e. ungroup addresses, retrieve information about IRIS and construct a CSV file.
+ """
+ log.info("... construction of dataset is in progress ...")
+
+ # 1. read data to transform
+ if not os.path.exists(self.old_path): clean()
+ data = pd.read_csv(self.old_path)
+
+ # 2. define some variables
+ raw_data = []
+ append_col = True
+ null_iris = 0 # count IRIS that can't be retrieve from its code
+ foreigners = 0 # count addresses that are not located in France
+ problems = 0 # count IRIS that encounter problems and that will be not added to the dataset
+ cols = []
+ columns = ["address", "country"]
+ columns.extend(ENVIRONMENT_VARIABLES)
+ cols_departure = [OLD_PREFIX+env for env in columns]
+ cols_arrival = [NEW_PREFIX+env for env in columns]
+
+ # 3. get data for departure and arrival IRIS
+ departures = data[cols_departure]
+ arrivals = data[cols_arrival]
+ # remove "old_address", "old_country", "new_address", "new_country"
+ cols_departure.pop(0)
+ cols_departure.pop(0)
+ cols_arrival.pop(0)
+ cols_arrival.pop(0)
+ cols.append("CODE")
+ cols.append("AREA")
+ cols.append("DENSITY")
+
+ # 4. for each departure and arrival IRIS (one line in cleanedData.csv), get its INSEE indicators and its EV.
+ # for each IRIS store its code, compute and store its area and its density, store all its INSEE indicators and its 6 EV.
+ for (index1, row1), (index2, row2) in zip(departures.iterrows(), arrivals.iterrows()):
+ dep_row, arr_row = [], []
+ # convert address to IRIS code and append it to data only if the address is in France
+ if row1[OLD_PREFIX + "country"] == "France":
+ dep_code = address_to_code(row1[OLD_PREFIX + "address"])
+ dep_row.append(dep_code)
+ dep_iris = model.get_iris_from_code(dep_code) # get IRIS information
+ coords_dep = model.get_coords_from_code(dep_code)
+ area_dep = area(coords_dep) / 1000000 if coords_dep is not None else None # convert area from m^2 to km^2
+ dep_row.append(area_dep) # add area as a feature
+ density = dep_iris["properties"]["raw_indicators"]["P14_POP"] / area_dep if area_dep is not None and area_dep > 0 else None
+ dep_row.append(density) # add density as a feature
+
+ if dep_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ dep_row, cols = append_indicator(raw_indicator, dep_iris, dep_row, append_col, cols)
+
+ # b. append EV
+ for target in cols_departure:
+ dep_row = append_target(row1, target, dep_row)
+ append_col = False
+ if len(dep_row) > 0: raw_data.append(dep_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row1[OLD_PREFIX+"country"] == "" or row1[OLD_PREFIX+"country"] is None or row1[OLD_PREFIX+"country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+ if row2[NEW_PREFIX + "country"] == "France":
+ arr_code = address_to_code(row2[NEW_PREFIX + "address"])
+ arr_row.append(arr_code)
+ arr_iris = model.get_iris_from_code(arr_code)
+ coords_arr = model.get_coords_from_code(arr_code)
+ area_arr = area(coords_arr) / 1000000 if coords_arr is not None else None # convert area from m^2 to km^2
+ arr_row.append(area_arr) # add area as a feature
+ density = arr_iris["properties"]["raw_indicators"]["P14_POP"] / area_arr if area_arr is not None and area_arr > 0 else None
+ arr_row.append(density) # add density as a feature
+
+ if arr_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ arr_row, cols = append_indicator(raw_indicator, arr_iris, arr_row, append_col, cols)
+
+ # b. append targets
+ for target in cols_arrival:
+ arr_row = append_target(row2, target, arr_row)
+ append_col = False
+ if len(arr_row) > 0: raw_data.append(arr_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row2[NEW_PREFIX + "country"] == "" or row2[NEW_PREFIX + "country"] is None or row2[NEW_PREFIX + "country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+
+ sys.stdout.write("\r%.2f%%" % ((index1 * 100) / len(departures))) # update progress percentage
+ sys.stdout.flush()
+ print()
+
+ cols.extend(ENVIRONMENT_VARIABLES)
+ log.info("%d addresses are not located in France.", foreigners)
+ log.info("%d null IRIS have been removed from the dataset.", null_iris)
+ log.info("%d IRIS have encountered problems.", problems)
+
+ # 5. convert IRIS data into a DataFrame, fill missing values and remove fully empty columns.
+ self.data = pd.DataFrame(raw_data, columns=cols)
+ self.data.sort_values("CODE", inplace=True)
+ self.fill_missing_values("median")
+ nan_columns = self.data.columns[self.data.isna().all()].tolist()
+ log.debug("There are %d NaN columns", len(nan_columns))
+ self.data.drop(nan_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in nan_columns] # remove names of NaN columns in the list of indicators
+
+ def fill_missing_values(self, method="median"):
+ """
+ Fill NaN values given the method.
+ Args:
+ method: A string corresponding tto the method for filling NaN values ("zero", "mean" or "median"). Default is median.
+ """
+ assert method in ["zero", "mean", "median"]
+ cols = ["CODE", "AREA", "DENSITY"]
+ cols.extend(ENVIRONMENT_VARIABLES)
+ for col in self.data.iteritems():
+ value = 0
+ col_name = col[0]
+ if col_name not in cols:
+ # fill missing INSEE indicators
+ if method == "zero":
+ value = 0
+ elif method == "mean":
+ value = self.data[col_name].mean()
+ elif method == "median":
+ value = self.data[col_name].median()
+ else:
+ value = np.mean(self.data[col_name])
+ elif col_name not in ["CODE", "AREA", "DENSITY"]:
+ # fill missing EV
+ if method == "zero":
+ env_values = {"building_type": "Housing estates",
+ "building_usage": "Other activities",
+ "landscape": "Urban",
+ "morphological_position": "Central",
+ "geographical_position": "Centre",
+ "social_class": "Lower"}
+ else: # method == "mean" or method == "median"
+ env_values = {"building_type": "Mixed",
+ "building_usage": "Other activities",
+ "landscape": "Green areas",
+ "morphological_position": "Urban",
+ "geographical_position": "Centre",
+ "social_class": "Middle"}
+ value = env_values[col_name]
+ # else: -> column is CODE or AREA or DENSITY -> do nothing
+ self.data[col_name].replace([np.nan], [value], inplace=True)
+
+ def filter_too_detailed_indicators(self):
+ """
+ Remove too detailed indicators from the dataset. The list of indicators is given by the file `regrouping.csv`.
+ """
+ regrouping = pd.read_csv(FILE_GROUPING, sep="\t")
+ status = regrouping.index[regrouping["STATUS"] == 1].tolist() # indicators that are an element of a subset
+ col_names = [regrouping.iloc[status]["INDICATOR"] for status in status]
+ # self.filtered_data = self.data[:] # copy by value, not by reference
+ counter1, counter2 = 0, 0
+ for column in col_names:
+ if column in self.data and column in self.indicators and column:
+ del self.data[column]
+ self.indicators.remove(column)
+ counter1 += 1
+ log.debug("%d indicators have been using regrouping.", counter1)
+
+ def apply_filtering(self):
+ """
+ Remove (filter) too detailed indicators and constant columns from the dataset.
+ """
+ if self.data is None:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()
+
+ # 1. remove indicators that have been defined as useless
+ self.filter_too_detailed_indicators()
+
+ # 2. remove constant columns, i.e. with a null variance
+ constant_columns = self.data.columns[self.data.nunique() == 1].tolist()
+ self.data.drop(constant_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in constant_columns]
+
+ self.data.to_csv(self.filtered_path, index=None) # index=None: avoid line numbering
+
+ def apply_normalization(self):
+ """
+ Normalize the dataset with the given method, i.e. None, "population" or "density". Default is "density".
+ """
+ assert self.normalization in [None, "population", "density"]
+ do_not_normalize = ["CODE", "AREA", "DENSITY", "P14_POP"]
+ do_not_normalize.extend(ENVIRONMENT_VARIABLES) # extend does not return a list
+ for index, row in self.data.iterrows():
+ for column in row.iteritems():
+ col_name = column[0]
+ if col_name not in do_not_normalize:
+ if self.normalization == "population":
+ self.data.at[index, col_name] = row[col_name] / row["P14_POP"]
+ elif self.normalization == "density":
+ density = row[self.data.columns.get_loc("DENSITY")]
+ self.data.at[index, col_name] = row[col_name] / density if density > 0 else 0
+ elif self.normalization is None:
+ self.data.at[index, col_name] = row[col_name]
+ self.data.to_csv(self.dataset_path, index=None) # index=None: avoid line numbering</code></pre>
+</details>
+<h3>Methods</h3>
+<dl>
+<dt id="Data.Data.apply_filtering"><code class="name flex">
+<span>def <span class="ident">apply_filtering</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Remove (filter) too detailed indicators and constant columns from the dataset.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def apply_filtering(self):
+ """
+ Remove (filter) too detailed indicators and constant columns from the dataset.
+ """
+ if self.data is None:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()
+
+ # 1. remove indicators that have been defined as useless
+ self.filter_too_detailed_indicators()
+
+ # 2. remove constant columns, i.e. with a null variance
+ constant_columns = self.data.columns[self.data.nunique() == 1].tolist()
+ self.data.drop(constant_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in constant_columns]
+
+ self.data.to_csv(self.filtered_path, index=None) # index=None: avoid line numbering</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.apply_normalization"><code class="name flex">
+<span>def <span class="ident">apply_normalization</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Normalize the dataset with the given method, i.e. None, "population" or "density". Default is "density".</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def apply_normalization(self):
+ """
+ Normalize the dataset with the given method, i.e. None, "population" or "density". Default is "density".
+ """
+ assert self.normalization in [None, "population", "density"]
+ do_not_normalize = ["CODE", "AREA", "DENSITY", "P14_POP"]
+ do_not_normalize.extend(ENVIRONMENT_VARIABLES) # extend does not return a list
+ for index, row in self.data.iterrows():
+ for column in row.iteritems():
+ col_name = column[0]
+ if col_name not in do_not_normalize:
+ if self.normalization == "population":
+ self.data.at[index, col_name] = row[col_name] / row["P14_POP"]
+ elif self.normalization == "density":
+ density = row[self.data.columns.get_loc("DENSITY")]
+ self.data.at[index, col_name] = row[col_name] / density if density > 0 else 0
+ elif self.normalization is None:
+ self.data.at[index, col_name] = row[col_name]
+ self.data.to_csv(self.dataset_path, index=None) # index=None: avoid line numbering</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.create"><code class="name flex">
+<span>def <span class="ident">create</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Construct the dataset from HomeInLove data, i.e. ungroup addresses, retrieve information about IRIS and construct a CSV file.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def create(self):
+ """
+ Construct the dataset from HomeInLove data, i.e. ungroup addresses, retrieve information about IRIS and construct a CSV file.
+ """
+ log.info("... construction of dataset is in progress ...")
+
+ # 1. read data to transform
+ if not os.path.exists(self.old_path): clean()
+ data = pd.read_csv(self.old_path)
+
+ # 2. define some variables
+ raw_data = []
+ append_col = True
+ null_iris = 0 # count IRIS that can't be retrieve from its code
+ foreigners = 0 # count addresses that are not located in France
+ problems = 0 # count IRIS that encounter problems and that will be not added to the dataset
+ cols = []
+ columns = ["address", "country"]
+ columns.extend(ENVIRONMENT_VARIABLES)
+ cols_departure = [OLD_PREFIX+env for env in columns]
+ cols_arrival = [NEW_PREFIX+env for env in columns]
+
+ # 3. get data for departure and arrival IRIS
+ departures = data[cols_departure]
+ arrivals = data[cols_arrival]
+ # remove "old_address", "old_country", "new_address", "new_country"
+ cols_departure.pop(0)
+ cols_departure.pop(0)
+ cols_arrival.pop(0)
+ cols_arrival.pop(0)
+ cols.append("CODE")
+ cols.append("AREA")
+ cols.append("DENSITY")
+
+ # 4. for each departure and arrival IRIS (one line in cleanedData.csv), get its INSEE indicators and its EV.
+ # for each IRIS store its code, compute and store its area and its density, store all its INSEE indicators and its 6 EV.
+ for (index1, row1), (index2, row2) in zip(departures.iterrows(), arrivals.iterrows()):
+ dep_row, arr_row = [], []
+ # convert address to IRIS code and append it to data only if the address is in France
+ if row1[OLD_PREFIX + "country"] == "France":
+ dep_code = address_to_code(row1[OLD_PREFIX + "address"])
+ dep_row.append(dep_code)
+ dep_iris = model.get_iris_from_code(dep_code) # get IRIS information
+ coords_dep = model.get_coords_from_code(dep_code)
+ area_dep = area(coords_dep) / 1000000 if coords_dep is not None else None # convert area from m^2 to km^2
+ dep_row.append(area_dep) # add area as a feature
+ density = dep_iris["properties"]["raw_indicators"]["P14_POP"] / area_dep if area_dep is not None and area_dep > 0 else None
+ dep_row.append(density) # add density as a feature
+
+ if dep_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ dep_row, cols = append_indicator(raw_indicator, dep_iris, dep_row, append_col, cols)
+
+ # b. append EV
+ for target in cols_departure:
+ dep_row = append_target(row1, target, dep_row)
+ append_col = False
+ if len(dep_row) > 0: raw_data.append(dep_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row1[OLD_PREFIX+"country"] == "" or row1[OLD_PREFIX+"country"] is None or row1[OLD_PREFIX+"country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+ if row2[NEW_PREFIX + "country"] == "France":
+ arr_code = address_to_code(row2[NEW_PREFIX + "address"])
+ arr_row.append(arr_code)
+ arr_iris = model.get_iris_from_code(arr_code)
+ coords_arr = model.get_coords_from_code(arr_code)
+ area_arr = area(coords_arr) / 1000000 if coords_arr is not None else None # convert area from m^2 to km^2
+ arr_row.append(area_arr) # add area as a feature
+ density = arr_iris["properties"]["raw_indicators"]["P14_POP"] / area_arr if area_arr is not None and area_arr > 0 else None
+ arr_row.append(density) # add density as a feature
+
+ if arr_iris is not None:
+ # a. append INSEE indicators
+ for raw_indicator in self.indicators:
+ arr_row, cols = append_indicator(raw_indicator, arr_iris, arr_row, append_col, cols)
+
+ # b. append targets
+ for target in cols_arrival:
+ arr_row = append_target(row2, target, arr_row)
+ append_col = False
+ if len(arr_row) > 0: raw_data.append(arr_row)
+ else: problems += 1
+ else: null_iris += 1
+ elif row2[NEW_PREFIX + "country"] == "" or row2[NEW_PREFIX + "country"] is None or row2[NEW_PREFIX + "country"] is np.nan: null_iris += 1
+ else: foreigners += 1
+
+ sys.stdout.write("\r%.2f%%" % ((index1 * 100) / len(departures))) # update progress percentage
+ sys.stdout.flush()
+ print()
+
+ cols.extend(ENVIRONMENT_VARIABLES)
+ log.info("%d addresses are not located in France.", foreigners)
+ log.info("%d null IRIS have been removed from the dataset.", null_iris)
+ log.info("%d IRIS have encountered problems.", problems)
+
+ # 5. convert IRIS data into a DataFrame, fill missing values and remove fully empty columns.
+ self.data = pd.DataFrame(raw_data, columns=cols)
+ self.data.sort_values("CODE", inplace=True)
+ self.fill_missing_values("median")
+ nan_columns = self.data.columns[self.data.isna().all()].tolist()
+ log.debug("There are %d NaN columns", len(nan_columns))
+ self.data.drop(nan_columns, axis=1, inplace=True)
+ self.indicators = [indicator for indicator in self.indicators if indicator not in nan_columns] # remove names of NaN columns in the list of indicators</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.fill_missing_values"><code class="name flex">
+<span>def <span class="ident">fill_missing_values</span></span>(<span>self, method='median')</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Fill NaN values given the method.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>method</code></strong></dt>
+<dd>A string corresponding tto the method for filling NaN values ("zero", "mean" or "median"). Default is median.</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def fill_missing_values(self, method="median"):
+ """
+ Fill NaN values given the method.
+ Args:
+ method: A string corresponding tto the method for filling NaN values ("zero", "mean" or "median"). Default is median.
+ """
+ assert method in ["zero", "mean", "median"]
+ cols = ["CODE", "AREA", "DENSITY"]
+ cols.extend(ENVIRONMENT_VARIABLES)
+ for col in self.data.iteritems():
+ value = 0
+ col_name = col[0]
+ if col_name not in cols:
+ # fill missing INSEE indicators
+ if method == "zero":
+ value = 0
+ elif method == "mean":
+ value = self.data[col_name].mean()
+ elif method == "median":
+ value = self.data[col_name].median()
+ else:
+ value = np.mean(self.data[col_name])
+ elif col_name not in ["CODE", "AREA", "DENSITY"]:
+ # fill missing EV
+ if method == "zero":
+ env_values = {"building_type": "Housing estates",
+ "building_usage": "Other activities",
+ "landscape": "Urban",
+ "morphological_position": "Central",
+ "geographical_position": "Centre",
+ "social_class": "Lower"}
+ else: # method == "mean" or method == "median"
+ env_values = {"building_type": "Mixed",
+ "building_usage": "Other activities",
+ "landscape": "Green areas",
+ "morphological_position": "Urban",
+ "geographical_position": "Centre",
+ "social_class": "Middle"}
+ value = env_values[col_name]
+ # else: -> column is CODE or AREA or DENSITY -> do nothing
+ self.data[col_name].replace([np.nan], [value], inplace=True)</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.filter_too_detailed_indicators"><code class="name flex">
+<span>def <span class="ident">filter_too_detailed_indicators</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Remove too detailed indicators from the dataset. The list of indicators is given by the file <code>regrouping.csv</code>.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def filter_too_detailed_indicators(self):
+ """
+ Remove too detailed indicators from the dataset. The list of indicators is given by the file `regrouping.csv`.
+ """
+ regrouping = pd.read_csv(FILE_GROUPING, sep="\t")
+ status = regrouping.index[regrouping["STATUS"] == 1].tolist() # indicators that are an element of a subset
+ col_names = [regrouping.iloc[status]["INDICATOR"] for status in status]
+ # self.filtered_data = self.data[:] # copy by value, not by reference
+ counter1, counter2 = 0, 0
+ for column in col_names:
+ if column in self.data and column in self.indicators and column:
+ del self.data[column]
+ self.indicators.remove(column)
+ counter1 += 1
+ log.debug("%d indicators have been using regrouping.", counter1)</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.get_indicators"><code class="name flex">
+<span>def <span class="ident">get_indicators</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get indicators from the dataset if it exists, else from the database.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_indicators(self):
+ """
+ Get indicators from the dataset if it exists, else from the database.
+ """
+ if self.data is not None:
+ self.indicators = self.data.columns.tolist()
+ # self.data.columns gets all columns, so to get indicators, we remove "CODE" (since it is not relevant for prediction).
+ # We keep "AREA" and "DENSITY" since they are use as features in prediction
+ if "CODE" in self.indicators: self.indicators.remove("CODE")
+ for env in ENVIRONMENT_VARIABLES:
+ if env in self.indicators:
+ self.indicators.remove(env)
+ else:
+ self.indicators = model.get_indicators_list()</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.init_all_in_one"><code class="name flex">
+<span>def <span class="ident">init_all_in_one</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Create or read the dataset. It applies normalization and/or filtering if needed.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def init_all_in_one(self):
+ """
+ Create or read the dataset. It applies normalization and/or filtering if needed.
+ """
+ if not os.path.exists(self.dataset_path):
+ self.create()
+ self.apply_normalization()
+ if self.filtering: self.apply_filtering()
+ elif not os.path.exists(self.filtered_path):
+ if self.filtering: self.apply_filtering()
+ else:
+ self.read()</code></pre>
+</details>
+</dd>
+<dt id="Data.Data.read"><code class="name flex">
+<span>def <span class="ident">read</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Read the dataset stored in the CSV file (and get indicators from this dataset).</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def read(self):
+ """
+ Read the dataset stored in the CSV file (and get indicators from this dataset).
+ """
+ if self.filtering:
+ self.data = pd.read_csv(self.filtered_path)
+ else:
+ self.data = pd.read_csv(self.dataset_path)
+ self.get_indicators()</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="Data.Data" href="#Data.Data">Data</a></code></h4>
+<ul class="">
+<li><code><a title="Data.Data.apply_filtering" href="#Data.Data.apply_filtering">apply_filtering</a></code></li>
+<li><code><a title="Data.Data.apply_normalization" href="#Data.Data.apply_normalization">apply_normalization</a></code></li>
+<li><code><a title="Data.Data.create" href="#Data.Data.create">create</a></code></li>
+<li><code><a title="Data.Data.fill_missing_values" href="#Data.Data.fill_missing_values">fill_missing_values</a></code></li>
+<li><code><a title="Data.Data.filter_too_detailed_indicators" href="#Data.Data.filter_too_detailed_indicators">filter_too_detailed_indicators</a></code></li>
+<li><code><a title="Data.Data.get_indicators" href="#Data.Data.get_indicators">get_indicators</a></code></li>
+<li><code><a title="Data.Data.init_all_in_one" href="#Data.Data.init_all_in_one">init_all_in_one</a></code></li>
+<li><code><a title="Data.Data.read" href="#Data.Data.read">read</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/Dataset.html b/doc/Dataset.html
new file mode 100644
index 00000000..bfedcc2b
--- /dev/null
+++ b/doc/Dataset.html
@@ -0,0 +1,515 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>Dataset API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>Dataset</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import pandas as pd
+
+from predihood.config import TRAIN_SIZE, TEST_SIZE, ENVIRONMENT_VARIABLES, RANDOM_STATE, FILE_ENV
+from predihood.utility_functions import check_dataset_size
+from sklearn.ensemble import IsolationForest
+from sklearn.model_selection import train_test_split
+
+log = logging.getLogger(__name__)
+
+
+class Dataset:
+ """
+ This class represents assessed IRIS with their indicators ans EV values. There are options, such as removing outliers or rural IRIS.
+ """
+ def __init__(self, data, env, _type, selected_indicators=None, indicators_to_remove=None, train_size=TRAIN_SIZE, test_size=TEST_SIZE, outliers=False):
+ """
+ Constructor of the Dataset class. Initialize attributes.
+
+ Args:
+ data: an instance of Data class. Don"t forget to initialize data after created it with "init_all_in_one()" method
+ env: a string representing the EV, i.e. a value in ["building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]
+ selected_indicators: a list containing the indicators to keep in the dataset
+ indicators_to_remove: a list containing the indicators to remove in the dataset
+ train_size: a integer or a float corresponding to the size of the dataset used for training
+ test_size: a integer or a float corresponding to the size of the dataset used for test
+ outliers: True or False to remove outliers from dataset (detected with IsolationForest algorithm)
+ """
+ self.data = data.data[:] # data must be a Data object
+ self.indicators = data.indicators[:]
+ self.filtering = data.filtering
+ self.normalization = data.normalization
+ self.selected_indicators = selected_indicators[:] if selected_indicators is not None else None
+ self.indicators_to_remove = indicators_to_remove[:] if indicators_to_remove is not None else None
+ self.type = _type
+ self.X = None
+ self.Y = None
+ self.X_train = None
+ self.Y_train = None
+ self.X_test = None
+ self.Y_test = None
+ if env in ENVIRONMENT_VARIABLES: self.env = env
+ else: self.env = "building_type"
+ self.train_size, self.test_size = check_dataset_size(train_size, test_size)
+ self.outliers = outliers
+
+ def init_all_in_one(self):
+ """
+ Initialize the dataset by initializing X and Y ; generating X_train, Y_train, X_test, Y_test ; removing outliers if needed.
+ When the type is "unsupervised", split data into X and Y is not relevant (as there is no train/test sets).
+ """
+ if self.type == "supervised":
+ if self.outliers:
+ self.remove_outliers()
+ self.init_X()
+ self.init_Y()
+ self.train_test()
+
+ def init_X(self):
+ """
+ Initialize self.X by getting indicators in dataset.
+ """
+ assert self.data is not None
+ if self.selected_indicators is not None: # select given indicators
+ self.X = self.data.loc[:, self.selected_indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.selected_indicators and indicator in self.X.columns:
+ self.selected_indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)
+ else:
+ self.X = self.data.loc[:, self.indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.indicators and indicator in self.X.columns:
+ self.indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)
+
+ def init_Y(self):
+ """
+ Initialize self.Y by getting EV in dataset.
+ """
+ assert self.data is not None
+ self.Y = self.data[self.env].values
+
+ def train_test(self):
+ """
+ Create X_train, Y_train, X_test, Y_test with train_test_split method
+ """
+ if len(self.X) <= 0: self.init_X()
+ if len(self.Y) <= 0: self.init_Y()
+ self.X_train, self.X_test, self.Y_train, self.Y_test = train_test_split(self.X, self.Y, train_size=self.train_size, test_size=self.test_size, random_state=RANDOM_STATE) # , stratify=self.Y)
+
+ def remove_outliers(self):
+ """
+ Detect and remove IRIS that are outliers from dataset.
+ """
+ isolation_forest = IsolationForest(random_state=RANDOM_STATE, n_estimators=100) # IsolationForest is used to detect outliers
+ isolation_forest.fit(self.X_train)
+ predictions = isolation_forest.predict(self.X_test)
+ for i in range(len(predictions)):
+ if predictions[i] == -1:
+ code = self.data.loc[i, "CODE"]
+ log.debug(code, "has been removed since it is an outlier.")
+ self.data = self.data.drop(i, axis=0) # delete IRIS that are detected as outliers
+
+ def remove_rural_iris(self):
+ """
+ Remove from dataset IRIS that are assessed as rural (because they bias the prediction).
+ """
+ self.data = self.data[self.data.morphological_position != "Rural"]
+
+ def get_environment_variable(self):
+ """
+ Get values for a given EV for each assessed IRIS and store it in a CSV file.
+ """
+ assert self.env in ENVIRONMENT_VARIABLES
+ data_env = pd.DataFrame(self.data[['CODE', self.env]])
+ data_env.to_csv(FILE_ENV)
+
+ def get_all_environmental_variable(self):
+ """
+ Get EV for each assessed IRIS and store it in a CSV file.
+ """
+ columns = ['CODE']
+ columns.extend(ENVIRONMENT_VARIABLES)
+ data_env = pd.DataFrame(self.data[columns])
+ data_env.to_csv(FILE_ENV)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="Dataset.Dataset"><code class="flex name class">
+<span>class <span class="ident">Dataset</span></span>
+<span>(</span><span>data, env, _type, selected_indicators=None, indicators_to_remove=None, train_size=0.8, test_size=0.2, outliers=False)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>This class represents assessed IRIS with their indicators ans EV values. There are options, such as removing outliers or rural IRIS.</p>
+<p>Constructor of the Dataset class. Initialize attributes.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>data</code></strong></dt>
+<dd>an instance of Data class. Don"t forget to initialize data after created it with "init_all_in_one()" method</dd>
+<dt><strong><code>env</code></strong></dt>
+<dd>a string representing the EV, i.e. a value in ["building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]</dd>
+<dt><strong><code>selected_indicators</code></strong></dt>
+<dd>a list containing the indicators to keep in the dataset</dd>
+<dt><strong><code>indicators_to_remove</code></strong></dt>
+<dd>a list containing the indicators to remove in the dataset</dd>
+<dt><strong><code>train_size</code></strong></dt>
+<dd>a integer or a float corresponding to the size of the dataset used for training</dd>
+<dt><strong><code>test_size</code></strong></dt>
+<dd>a integer or a float corresponding to the size of the dataset used for test</dd>
+<dt><strong><code>outliers</code></strong></dt>
+<dd>True or False to remove outliers from dataset (detected with IsolationForest algorithm)</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class Dataset:
+ """
+ This class represents assessed IRIS with their indicators ans EV values. There are options, such as removing outliers or rural IRIS.
+ """
+ def __init__(self, data, env, _type, selected_indicators=None, indicators_to_remove=None, train_size=TRAIN_SIZE, test_size=TEST_SIZE, outliers=False):
+ """
+ Constructor of the Dataset class. Initialize attributes.
+
+ Args:
+ data: an instance of Data class. Don"t forget to initialize data after created it with "init_all_in_one()" method
+ env: a string representing the EV, i.e. a value in ["building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]
+ selected_indicators: a list containing the indicators to keep in the dataset
+ indicators_to_remove: a list containing the indicators to remove in the dataset
+ train_size: a integer or a float corresponding to the size of the dataset used for training
+ test_size: a integer or a float corresponding to the size of the dataset used for test
+ outliers: True or False to remove outliers from dataset (detected with IsolationForest algorithm)
+ """
+ self.data = data.data[:] # data must be a Data object
+ self.indicators = data.indicators[:]
+ self.filtering = data.filtering
+ self.normalization = data.normalization
+ self.selected_indicators = selected_indicators[:] if selected_indicators is not None else None
+ self.indicators_to_remove = indicators_to_remove[:] if indicators_to_remove is not None else None
+ self.type = _type
+ self.X = None
+ self.Y = None
+ self.X_train = None
+ self.Y_train = None
+ self.X_test = None
+ self.Y_test = None
+ if env in ENVIRONMENT_VARIABLES: self.env = env
+ else: self.env = "building_type"
+ self.train_size, self.test_size = check_dataset_size(train_size, test_size)
+ self.outliers = outliers
+
+ def init_all_in_one(self):
+ """
+ Initialize the dataset by initializing X and Y ; generating X_train, Y_train, X_test, Y_test ; removing outliers if needed.
+ When the type is "unsupervised", split data into X and Y is not relevant (as there is no train/test sets).
+ """
+ if self.type == "supervised":
+ if self.outliers:
+ self.remove_outliers()
+ self.init_X()
+ self.init_Y()
+ self.train_test()
+
+ def init_X(self):
+ """
+ Initialize self.X by getting indicators in dataset.
+ """
+ assert self.data is not None
+ if self.selected_indicators is not None: # select given indicators
+ self.X = self.data.loc[:, self.selected_indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.selected_indicators and indicator in self.X.columns:
+ self.selected_indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)
+ else:
+ self.X = self.data.loc[:, self.indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.indicators and indicator in self.X.columns:
+ self.indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)
+
+ def init_Y(self):
+ """
+ Initialize self.Y by getting EV in dataset.
+ """
+ assert self.data is not None
+ self.Y = self.data[self.env].values
+
+ def train_test(self):
+ """
+ Create X_train, Y_train, X_test, Y_test with train_test_split method
+ """
+ if len(self.X) <= 0: self.init_X()
+ if len(self.Y) <= 0: self.init_Y()
+ self.X_train, self.X_test, self.Y_train, self.Y_test = train_test_split(self.X, self.Y, train_size=self.train_size, test_size=self.test_size, random_state=RANDOM_STATE) # , stratify=self.Y)
+
+ def remove_outliers(self):
+ """
+ Detect and remove IRIS that are outliers from dataset.
+ """
+ isolation_forest = IsolationForest(random_state=RANDOM_STATE, n_estimators=100) # IsolationForest is used to detect outliers
+ isolation_forest.fit(self.X_train)
+ predictions = isolation_forest.predict(self.X_test)
+ for i in range(len(predictions)):
+ if predictions[i] == -1:
+ code = self.data.loc[i, "CODE"]
+ log.debug(code, "has been removed since it is an outlier.")
+ self.data = self.data.drop(i, axis=0) # delete IRIS that are detected as outliers
+
+ def remove_rural_iris(self):
+ """
+ Remove from dataset IRIS that are assessed as rural (because they bias the prediction).
+ """
+ self.data = self.data[self.data.morphological_position != "Rural"]
+
+ def get_environment_variable(self):
+ """
+ Get values for a given EV for each assessed IRIS and store it in a CSV file.
+ """
+ assert self.env in ENVIRONMENT_VARIABLES
+ data_env = pd.DataFrame(self.data[['CODE', self.env]])
+ data_env.to_csv(FILE_ENV)
+
+ def get_all_environmental_variable(self):
+ """
+ Get EV for each assessed IRIS and store it in a CSV file.
+ """
+ columns = ['CODE']
+ columns.extend(ENVIRONMENT_VARIABLES)
+ data_env = pd.DataFrame(self.data[columns])
+ data_env.to_csv(FILE_ENV)</code></pre>
+</details>
+<h3>Methods</h3>
+<dl>
+<dt id="Dataset.Dataset.get_all_environmental_variable"><code class="name flex">
+<span>def <span class="ident">get_all_environmental_variable</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get EV for each assessed IRIS and store it in a CSV file.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_all_environmental_variable(self):
+ """
+ Get EV for each assessed IRIS and store it in a CSV file.
+ """
+ columns = ['CODE']
+ columns.extend(ENVIRONMENT_VARIABLES)
+ data_env = pd.DataFrame(self.data[columns])
+ data_env.to_csv(FILE_ENV)</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.get_environment_variable"><code class="name flex">
+<span>def <span class="ident">get_environment_variable</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get values for a given EV for each assessed IRIS and store it in a CSV file.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_environment_variable(self):
+ """
+ Get values for a given EV for each assessed IRIS and store it in a CSV file.
+ """
+ assert self.env in ENVIRONMENT_VARIABLES
+ data_env = pd.DataFrame(self.data[['CODE', self.env]])
+ data_env.to_csv(FILE_ENV)</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.init_X"><code class="name flex">
+<span>def <span class="ident">init_X</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Initialize self.X by getting indicators in dataset.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def init_X(self):
+ """
+ Initialize self.X by getting indicators in dataset.
+ """
+ assert self.data is not None
+ if self.selected_indicators is not None: # select given indicators
+ self.X = self.data.loc[:, self.selected_indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.selected_indicators and indicator in self.X.columns:
+ self.selected_indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)
+ else:
+ self.X = self.data.loc[:, self.indicators]
+ if self.indicators_to_remove: # remove given indicators
+ for indicator in self.indicators_to_remove:
+ if indicator in self.indicators and indicator in self.X.columns:
+ self.indicators.remove(indicator)
+ self.X = self.X.drop([indicator], axis=1)</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.init_Y"><code class="name flex">
+<span>def <span class="ident">init_Y</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Initialize self.Y by getting EV in dataset.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def init_Y(self):
+ """
+ Initialize self.Y by getting EV in dataset.
+ """
+ assert self.data is not None
+ self.Y = self.data[self.env].values</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.init_all_in_one"><code class="name flex">
+<span>def <span class="ident">init_all_in_one</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Initialize the dataset by initializing X and Y ; generating X_train, Y_train, X_test, Y_test ; removing outliers if needed.
+When the type is "unsupervised", split data into X and Y is not relevant (as there is no train/test sets).</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def init_all_in_one(self):
+ """
+ Initialize the dataset by initializing X and Y ; generating X_train, Y_train, X_test, Y_test ; removing outliers if needed.
+ When the type is "unsupervised", split data into X and Y is not relevant (as there is no train/test sets).
+ """
+ if self.type == "supervised":
+ if self.outliers:
+ self.remove_outliers()
+ self.init_X()
+ self.init_Y()
+ self.train_test()</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.remove_outliers"><code class="name flex">
+<span>def <span class="ident">remove_outliers</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Detect and remove IRIS that are outliers from dataset.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def remove_outliers(self):
+ """
+ Detect and remove IRIS that are outliers from dataset.
+ """
+ isolation_forest = IsolationForest(random_state=RANDOM_STATE, n_estimators=100) # IsolationForest is used to detect outliers
+ isolation_forest.fit(self.X_train)
+ predictions = isolation_forest.predict(self.X_test)
+ for i in range(len(predictions)):
+ if predictions[i] == -1:
+ code = self.data.loc[i, "CODE"]
+ log.debug(code, "has been removed since it is an outlier.")
+ self.data = self.data.drop(i, axis=0) # delete IRIS that are detected as outliers</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.remove_rural_iris"><code class="name flex">
+<span>def <span class="ident">remove_rural_iris</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Remove from dataset IRIS that are assessed as rural (because they bias the prediction).</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def remove_rural_iris(self):
+ """
+ Remove from dataset IRIS that are assessed as rural (because they bias the prediction).
+ """
+ self.data = self.data[self.data.morphological_position != "Rural"]</code></pre>
+</details>
+</dd>
+<dt id="Dataset.Dataset.train_test"><code class="name flex">
+<span>def <span class="ident">train_test</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Create X_train, Y_train, X_test, Y_test with train_test_split method</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def train_test(self):
+ """
+ Create X_train, Y_train, X_test, Y_test with train_test_split method
+ """
+ if len(self.X) <= 0: self.init_X()
+ if len(self.Y) <= 0: self.init_Y()
+ self.X_train, self.X_test, self.Y_train, self.Y_test = train_test_split(self.X, self.Y, train_size=self.train_size, test_size=self.test_size, random_state=RANDOM_STATE) # , stratify=self.Y)</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="Dataset.Dataset" href="#Dataset.Dataset">Dataset</a></code></h4>
+<ul class="">
+<li><code><a title="Dataset.Dataset.get_all_environmental_variable" href="#Dataset.Dataset.get_all_environmental_variable">get_all_environmental_variable</a></code></li>
+<li><code><a title="Dataset.Dataset.get_environment_variable" href="#Dataset.Dataset.get_environment_variable">get_environment_variable</a></code></li>
+<li><code><a title="Dataset.Dataset.init_X" href="#Dataset.Dataset.init_X">init_X</a></code></li>
+<li><code><a title="Dataset.Dataset.init_Y" href="#Dataset.Dataset.init_Y">init_Y</a></code></li>
+<li><code><a title="Dataset.Dataset.init_all_in_one" href="#Dataset.Dataset.init_all_in_one">init_all_in_one</a></code></li>
+<li><code><a title="Dataset.Dataset.remove_outliers" href="#Dataset.Dataset.remove_outliers">remove_outliers</a></code></li>
+<li><code><a title="Dataset.Dataset.remove_rural_iris" href="#Dataset.Dataset.remove_rural_iris">remove_rural_iris</a></code></li>
+<li><code><a title="Dataset.Dataset.train_test" href="#Dataset.Dataset.train_test">train_test</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/Method.html b/doc/Method.html
new file mode 100644
index 00000000..bf7918a9
--- /dev/null
+++ b/doc/Method.html
@@ -0,0 +1,178 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>Method API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>Method</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">from predihood.config import RANDOM_STATE
+
+
+class Method:
+ """
+ This class represents the general concept of a Method that is applied on data. There are two specific concepts of Method: MethodSelection and MethodPrediction.
+ """
+ def __init__(self, name, dataset=None, classifier=None):
+ """
+ Constructor of the Method class. Initialize attributes.
+ Args:
+ name: a string that represents the name of the method, e.g. "feature selection" or "correlation matrix"
+ dataset: a Dataset object on which the method will be applied
+ classifier: an object that can be used (e.g. fit) on data
+ """
+ self.name = name
+ self.dataset = dataset
+ self.classifier = classifier
+ self.parameters = None # same as return
+
+ def fit(self):
+ """
+ Fit the classifier on dataset.
+ """
+ self.classifier.random_state = RANDOM_STATE # defined random_state to have reproducibility
+ if self.dataset.type == "supervised":
+ self.classifier.fit(X=self.dataset.X_train, y=self.dataset.Y_train)
+ else:
+ # unsupervised learning does not need split between X and Y
+ if self.dataset.selected_indicators is not None:
+ self.classifier.fit(self.dataset.data[self.dataset.selected_indicators])
+ else:
+ self.classifier.fit(self.dataset.data[self.dataset.indicators])</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="Method.Method"><code class="flex name class">
+<span>class <span class="ident">Method</span></span>
+<span>(</span><span>name, dataset=None, classifier=None)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>This class represents the general concept of a Method that is applied on data. There are two specific concepts of Method: MethodSelection and MethodPrediction.</p>
+<p>Constructor of the Method class. Initialize attributes.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>name</code></strong></dt>
+<dd>a string that represents the name of the method, e.g. "feature selection" or "correlation matrix"</dd>
+<dt><strong><code>dataset</code></strong></dt>
+<dd>a Dataset object on which the method will be applied</dd>
+<dt><strong><code>classifier</code></strong></dt>
+<dd>an object that can be used (e.g. fit) on data</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class Method:
+ """
+ This class represents the general concept of a Method that is applied on data. There are two specific concepts of Method: MethodSelection and MethodPrediction.
+ """
+ def __init__(self, name, dataset=None, classifier=None):
+ """
+ Constructor of the Method class. Initialize attributes.
+ Args:
+ name: a string that represents the name of the method, e.g. "feature selection" or "correlation matrix"
+ dataset: a Dataset object on which the method will be applied
+ classifier: an object that can be used (e.g. fit) on data
+ """
+ self.name = name
+ self.dataset = dataset
+ self.classifier = classifier
+ self.parameters = None # same as return
+
+ def fit(self):
+ """
+ Fit the classifier on dataset.
+ """
+ self.classifier.random_state = RANDOM_STATE # defined random_state to have reproducibility
+ if self.dataset.type == "supervised":
+ self.classifier.fit(X=self.dataset.X_train, y=self.dataset.Y_train)
+ else:
+ # unsupervised learning does not need split between X and Y
+ if self.dataset.selected_indicators is not None:
+ self.classifier.fit(self.dataset.data[self.dataset.selected_indicators])
+ else:
+ self.classifier.fit(self.dataset.data[self.dataset.indicators])</code></pre>
+</details>
+<h3>Methods</h3>
+<dl>
+<dt id="Method.Method.fit"><code class="name flex">
+<span>def <span class="ident">fit</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Fit the classifier on dataset.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def fit(self):
+ """
+ Fit the classifier on dataset.
+ """
+ self.classifier.random_state = RANDOM_STATE # defined random_state to have reproducibility
+ if self.dataset.type == "supervised":
+ self.classifier.fit(X=self.dataset.X_train, y=self.dataset.Y_train)
+ else:
+ # unsupervised learning does not need split between X and Y
+ if self.dataset.selected_indicators is not None:
+ self.classifier.fit(self.dataset.data[self.dataset.selected_indicators])
+ else:
+ self.classifier.fit(self.dataset.data[self.dataset.indicators])</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="Method.Method" href="#Method.Method">Method</a></code></h4>
+<ul class="">
+<li><code><a title="Method.Method.fit" href="#Method.Method.fit">fit</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/MethodCleaning.html b/doc/MethodCleaning.html
new file mode 100644
index 00000000..c46676df
--- /dev/null
+++ b/doc/MethodCleaning.html
@@ -0,0 +1,705 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>MethodCleaning API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>MethodCleaning</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import os
+import warnings
+import logging
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+
+from predihood.classes.Method import Method
+from predihood.config import ENVIRONMENT_VARIABLES, FILE_CLEANED_DATA, FOLDER_DISTRIBUTION, OLD_PREFIX, NEW_PREFIX
+from predihood.utility_functions import similarity, auto_label
+
+log = logging.getLogger(__name__)
+warnings.simplefilter(action='ignore', category=FutureWarning)
+
+
+class MethodCleaning(Method):
+ """
+ This class represents the method for cleaning data given by the French company.
+ """
+ def __init__(self, name, dataset):
+ """
+ Constructor of the MethodCleansing class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset)
+ self.values_by_env = {} # dictionary to store values for each EV, e.g. [House, Houses]
+ self.columns_dep = [] # departure columns, e.g. old_building_type, old_building_usage,...
+ self.columns_arr = [] # arrival columns, e.g. new_building_type, new_building_usage, ...
+ self.columns = {
+ "occupation": {
+ "name_dep": "old_occupation",
+ "data_dep": self.dataset.old_occupation,
+ "name_arr": "new_occupation",
+ "data_arr": self.dataset.new_occupation
+ }
+ }
+
+ for env in ENVIRONMENT_VARIABLES:
+ temp = {
+ "name_dep": OLD_PREFIX + env,
+ "name_arr": NEW_PREFIX + env,
+ "data_dep": self.dataset[OLD_PREFIX + str(env)],
+ "data_arr": self.dataset[NEW_PREFIX + str(env)]
+ }
+ self.columns[env] = temp
+
+ for env in self.columns: self.columns_dep.append(self.columns[env]["data_dep"])
+ for env in self.columns: self.columns_arr.append(self.columns[env]["data_arr"])
+
+ # define outliers (to be removed)
+ self.outliers = ['Oui', 'Moyenne-sup', 'Location']
+
+ # plot variables
+ self.before = {} # departure values for each EV
+ self.after = {} # arrival values for each EV
+ self.labels = {} # labels for EV, e.g. urban, green areas, forest and country-side for the landscape variable
+
+ def clean(self):
+ """
+ Clean data from bad naming conventions. The idea of this function is to create a dictionary with all spellings for each value of each EV, e.g. [["Houses", "House"], ["Upper middle", "upper middle", "upper midddle"], ["Green areas"], ["Countryside"]].
+ This dictionary is constructed by computing similarities between each values and store each spelling and finally let the user choose the best one.
+ """
+ #
+ log.info("The data needs to be cleaned. For each list, write the correct word. For each EV, you will get its number of corrections and its error rate.")
+ # 1. getting wrong values in a dictionary ordered by env variable
+ self.values_by_env = {}
+ for col_dep, col_arr in zip(self.columns_dep, self.columns_arr):
+ col_name = col_dep.name[len(OLD_PREFIX):]
+ self.values_by_env[col_name] = []
+ for val in col_dep.unique(): # get possible values for the current column
+ index = similarity(val, self.values_by_env[col_name])
+ # if the value is similar to another, add it, else create an new array with it
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+ for val in col_arr.unique():
+ index = similarity(val, self.values_by_env[col_name])
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+
+ # 2. renaming these wrong values in data
+ for key, value in self.values_by_env.items():
+ col_name_old = OLD_PREFIX + key
+ col_name_new = NEW_PREFIX + key
+ nb_replacement_dep = 0
+ nb_replacement_arr = 0
+ for i in range(len(value)):
+ if len(value[i]) > 1:
+ arr_without_duplicates = list(dict.fromkeys(value[i]))
+ chosen_label = input(str(arr_without_duplicates) + ": ")
+ for label in value[i]:
+ if label != chosen_label: # if label == chosen_label: skip it because no replacement is needed
+ nb_replacement_dep += pd.Series(self.dataset[col_name_old] == label).sum()
+ nb_replacement_arr += pd.Series(self.dataset[col_name_new] == label).sum()
+ self.dataset.loc[self.dataset[col_name_old] == label, col_name_old] = chosen_label
+ self.dataset.loc[self.dataset[col_name_new] == label, col_name_new] = chosen_label
+ size = int(self.dataset.count()[OLD_PREFIX + key]) + int(self.dataset.count()[NEW_PREFIX + key])
+ mean_error = ((nb_replacement_dep + nb_replacement_arr) / size) * 100
+ log.debug("%d IRIS have been corrected for the EV %s, corresponding to an error rate of %.0f %%", (nb_replacement_dep + nb_replacement_arr), key, mean_error)
+
+ # 3. removing outliers from data
+ count = 0
+ for outlier in self.outliers:
+ self.dataset.drop(self.dataset[self.dataset.eq(outlier).any(1)].index, inplace=True)
+ count += 1
+ log.debug("%d incorrect values removed", count)
+
+ # 4. save data
+ self.dataset.to_csv(FILE_CLEANED_DATA, index=False, encoding='utf-8')
+ log.info("Cleaned data is in %s", FILE_CLEANED_DATA)
+
+ def create_before_after_labels(self, name_dep, name_arr):
+ """
+ Creates the lists 'before', 'after' and 'labels' from data.
+
+ Args:
+ name_dep: a string containing the name of the departure column, e.g. old_building_type, old_building_usage...
+ name_arr: a string containing the name of the arrival column, e.g. new_building_type, new_building_usage...
+ """
+ all_repartition = {}
+ self.before = {}
+ self.after = {}
+
+ for status, value in self.dataset[name_dep].value_counts().items():
+ if name_dep == OLD_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.before:
+ self.before[status] += value
+ else:
+ self.before[status] = value
+ else:
+ self.before[status] = value # self.dataset[values_before].value_counts()[status]
+
+ for status, value in self.dataset[name_arr].value_counts().items():
+ if name_arr == NEW_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.after:
+ self.after[status] += value
+ else:
+ self.after[status] = value
+ else:
+ self.after[status] = value # self.dataset[values_after].value_counts()[status]
+
+ # 2. merge before and after data in the same dictionary
+ for status in self.before:
+ all_repartition[status] = [self.before[status], 0]
+ for status in self.after:
+ if status not in all_repartition:
+ all_repartition[status] = [0, self.after[status]]
+ else:
+ all_repartition[status][1] = self.after[status]
+
+ # 3. convert dictionary in 3 arrays
+ self.before = []
+ self.after = []
+ self.labels = []
+ for key in all_repartition:
+ if not isinstance(key, float): # to remove nan values
+ self.before.append(all_repartition[key][0])
+ self.after.append(all_repartition[key][1])
+ self.labels.append(key)
+
+ def create_bar_chart(self, name, title):
+ """
+ Plot before/after charts.
+
+ Args:
+ name: a string containing the name of the EV to plot, e.g. building_type, building_usage, landscape, ...
+ title: a string containing the title of the plot
+ """
+ x = np.arange(len(self.labels)) # the label locations
+ width = 0.35
+
+ fig, ax = plt.subplots()
+
+ ax.bar(x - width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ bef = ax.bar(x - width / 2, self.before, width=width, label='Avant') # before data
+ ax.bar(x + width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ aft = ax.bar(x + width / 2, self.after, width=width, label='Après') # after data
+
+ ax.set_ylabel('Nombre de personnes')
+ plt.xticks(x, self.labels, rotation='vertical')
+ auto_label(bef, ax)
+ auto_label(aft, ax)
+ plt.tight_layout()
+ ax.legend()
+ filename = os.path.join(FOLDER_DISTRIBUTION, "distribution_" + name + ".png")
+ fig.savefig(filename)
+ ax.set_title(title)
+ plt.show()
+
+ def to_chart(self, env, name, title):
+ """
+ Create before/after data and plot it.
+ :param env:
+ :param name:
+ :param title:
+ Args:
+ env: a string containing the EV to plot, e.g. building_type, building_usage, landscape...
+ name: a string containing the name to save the file
+ title: a string containing the title of the plot
+ """
+ self.create_before_after_labels(self.columns[env]["name_dep"], self.columns[env]["name_arr"])
+ self.create_bar_chart(name, title)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="MethodCleaning.MethodCleaning"><code class="flex name class">
+<span>class <span class="ident">MethodCleaning</span></span>
+<span>(</span><span>name, dataset)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>This class represents the method for cleaning data given by the French company.</p>
+<p>Constructor of the MethodCleansing class. Initialize attributes.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class MethodCleaning(Method):
+ """
+ This class represents the method for cleaning data given by the French company.
+ """
+ def __init__(self, name, dataset):
+ """
+ Constructor of the MethodCleansing class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset)
+ self.values_by_env = {} # dictionary to store values for each EV, e.g. [House, Houses]
+ self.columns_dep = [] # departure columns, e.g. old_building_type, old_building_usage,...
+ self.columns_arr = [] # arrival columns, e.g. new_building_type, new_building_usage, ...
+ self.columns = {
+ "occupation": {
+ "name_dep": "old_occupation",
+ "data_dep": self.dataset.old_occupation,
+ "name_arr": "new_occupation",
+ "data_arr": self.dataset.new_occupation
+ }
+ }
+
+ for env in ENVIRONMENT_VARIABLES:
+ temp = {
+ "name_dep": OLD_PREFIX + env,
+ "name_arr": NEW_PREFIX + env,
+ "data_dep": self.dataset[OLD_PREFIX + str(env)],
+ "data_arr": self.dataset[NEW_PREFIX + str(env)]
+ }
+ self.columns[env] = temp
+
+ for env in self.columns: self.columns_dep.append(self.columns[env]["data_dep"])
+ for env in self.columns: self.columns_arr.append(self.columns[env]["data_arr"])
+
+ # define outliers (to be removed)
+ self.outliers = ['Oui', 'Moyenne-sup', 'Location']
+
+ # plot variables
+ self.before = {} # departure values for each EV
+ self.after = {} # arrival values for each EV
+ self.labels = {} # labels for EV, e.g. urban, green areas, forest and country-side for the landscape variable
+
+ def clean(self):
+ """
+ Clean data from bad naming conventions. The idea of this function is to create a dictionary with all spellings for each value of each EV, e.g. [["Houses", "House"], ["Upper middle", "upper middle", "upper midddle"], ["Green areas"], ["Countryside"]].
+ This dictionary is constructed by computing similarities between each values and store each spelling and finally let the user choose the best one.
+ """
+ #
+ log.info("The data needs to be cleaned. For each list, write the correct word. For each EV, you will get its number of corrections and its error rate.")
+ # 1. getting wrong values in a dictionary ordered by env variable
+ self.values_by_env = {}
+ for col_dep, col_arr in zip(self.columns_dep, self.columns_arr):
+ col_name = col_dep.name[len(OLD_PREFIX):]
+ self.values_by_env[col_name] = []
+ for val in col_dep.unique(): # get possible values for the current column
+ index = similarity(val, self.values_by_env[col_name])
+ # if the value is similar to another, add it, else create an new array with it
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+ for val in col_arr.unique():
+ index = similarity(val, self.values_by_env[col_name])
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+
+ # 2. renaming these wrong values in data
+ for key, value in self.values_by_env.items():
+ col_name_old = OLD_PREFIX + key
+ col_name_new = NEW_PREFIX + key
+ nb_replacement_dep = 0
+ nb_replacement_arr = 0
+ for i in range(len(value)):
+ if len(value[i]) > 1:
+ arr_without_duplicates = list(dict.fromkeys(value[i]))
+ chosen_label = input(str(arr_without_duplicates) + ": ")
+ for label in value[i]:
+ if label != chosen_label: # if label == chosen_label: skip it because no replacement is needed
+ nb_replacement_dep += pd.Series(self.dataset[col_name_old] == label).sum()
+ nb_replacement_arr += pd.Series(self.dataset[col_name_new] == label).sum()
+ self.dataset.loc[self.dataset[col_name_old] == label, col_name_old] = chosen_label
+ self.dataset.loc[self.dataset[col_name_new] == label, col_name_new] = chosen_label
+ size = int(self.dataset.count()[OLD_PREFIX + key]) + int(self.dataset.count()[NEW_PREFIX + key])
+ mean_error = ((nb_replacement_dep + nb_replacement_arr) / size) * 100
+ log.debug("%d IRIS have been corrected for the EV %s, corresponding to an error rate of %.0f %%", (nb_replacement_dep + nb_replacement_arr), key, mean_error)
+
+ # 3. removing outliers from data
+ count = 0
+ for outlier in self.outliers:
+ self.dataset.drop(self.dataset[self.dataset.eq(outlier).any(1)].index, inplace=True)
+ count += 1
+ log.debug("%d incorrect values removed", count)
+
+ # 4. save data
+ self.dataset.to_csv(FILE_CLEANED_DATA, index=False, encoding='utf-8')
+ log.info("Cleaned data is in %s", FILE_CLEANED_DATA)
+
+ def create_before_after_labels(self, name_dep, name_arr):
+ """
+ Creates the lists 'before', 'after' and 'labels' from data.
+
+ Args:
+ name_dep: a string containing the name of the departure column, e.g. old_building_type, old_building_usage...
+ name_arr: a string containing the name of the arrival column, e.g. new_building_type, new_building_usage...
+ """
+ all_repartition = {}
+ self.before = {}
+ self.after = {}
+
+ for status, value in self.dataset[name_dep].value_counts().items():
+ if name_dep == OLD_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.before:
+ self.before[status] += value
+ else:
+ self.before[status] = value
+ else:
+ self.before[status] = value # self.dataset[values_before].value_counts()[status]
+
+ for status, value in self.dataset[name_arr].value_counts().items():
+ if name_arr == NEW_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.after:
+ self.after[status] += value
+ else:
+ self.after[status] = value
+ else:
+ self.after[status] = value # self.dataset[values_after].value_counts()[status]
+
+ # 2. merge before and after data in the same dictionary
+ for status in self.before:
+ all_repartition[status] = [self.before[status], 0]
+ for status in self.after:
+ if status not in all_repartition:
+ all_repartition[status] = [0, self.after[status]]
+ else:
+ all_repartition[status][1] = self.after[status]
+
+ # 3. convert dictionary in 3 arrays
+ self.before = []
+ self.after = []
+ self.labels = []
+ for key in all_repartition:
+ if not isinstance(key, float): # to remove nan values
+ self.before.append(all_repartition[key][0])
+ self.after.append(all_repartition[key][1])
+ self.labels.append(key)
+
+ def create_bar_chart(self, name, title):
+ """
+ Plot before/after charts.
+
+ Args:
+ name: a string containing the name of the EV to plot, e.g. building_type, building_usage, landscape, ...
+ title: a string containing the title of the plot
+ """
+ x = np.arange(len(self.labels)) # the label locations
+ width = 0.35
+
+ fig, ax = plt.subplots()
+
+ ax.bar(x - width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ bef = ax.bar(x - width / 2, self.before, width=width, label='Avant') # before data
+ ax.bar(x + width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ aft = ax.bar(x + width / 2, self.after, width=width, label='Après') # after data
+
+ ax.set_ylabel('Nombre de personnes')
+ plt.xticks(x, self.labels, rotation='vertical')
+ auto_label(bef, ax)
+ auto_label(aft, ax)
+ plt.tight_layout()
+ ax.legend()
+ filename = os.path.join(FOLDER_DISTRIBUTION, "distribution_" + name + ".png")
+ fig.savefig(filename)
+ ax.set_title(title)
+ plt.show()
+
+ def to_chart(self, env, name, title):
+ """
+ Create before/after data and plot it.
+ :param env:
+ :param name:
+ :param title:
+ Args:
+ env: a string containing the EV to plot, e.g. building_type, building_usage, landscape...
+ name: a string containing the name to save the file
+ title: a string containing the title of the plot
+ """
+ self.create_before_after_labels(self.columns[env]["name_dep"], self.columns[env]["name_arr"])
+ self.create_bar_chart(name, title)</code></pre>
+</details>
+<h3>Ancestors</h3>
+<ul class="hlist">
+<li>predihood.classes.Method.Method</li>
+</ul>
+<h3>Methods</h3>
+<dl>
+<dt id="MethodCleaning.MethodCleaning.clean"><code class="name flex">
+<span>def <span class="ident">clean</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Clean data from bad naming conventions. The idea of this function is to create a dictionary with all spellings for each value of each EV, e.g. [["Houses", "House"], ["Upper middle", "upper middle", "upper midddle"], ["Green areas"], ["Countryside"]].
+This dictionary is constructed by computing similarities between each values and store each spelling and finally let the user choose the best one.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def clean(self):
+ """
+ Clean data from bad naming conventions. The idea of this function is to create a dictionary with all spellings for each value of each EV, e.g. [["Houses", "House"], ["Upper middle", "upper middle", "upper midddle"], ["Green areas"], ["Countryside"]].
+ This dictionary is constructed by computing similarities between each values and store each spelling and finally let the user choose the best one.
+ """
+ #
+ log.info("The data needs to be cleaned. For each list, write the correct word. For each EV, you will get its number of corrections and its error rate.")
+ # 1. getting wrong values in a dictionary ordered by env variable
+ self.values_by_env = {}
+ for col_dep, col_arr in zip(self.columns_dep, self.columns_arr):
+ col_name = col_dep.name[len(OLD_PREFIX):]
+ self.values_by_env[col_name] = []
+ for val in col_dep.unique(): # get possible values for the current column
+ index = similarity(val, self.values_by_env[col_name])
+ # if the value is similar to another, add it, else create an new array with it
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+ for val in col_arr.unique():
+ index = similarity(val, self.values_by_env[col_name])
+ if index >= 0:
+ self.values_by_env[col_name][index].append(val)
+ elif index == -1:
+ self.values_by_env[col_name].append([val])
+
+ # 2. renaming these wrong values in data
+ for key, value in self.values_by_env.items():
+ col_name_old = OLD_PREFIX + key
+ col_name_new = NEW_PREFIX + key
+ nb_replacement_dep = 0
+ nb_replacement_arr = 0
+ for i in range(len(value)):
+ if len(value[i]) > 1:
+ arr_without_duplicates = list(dict.fromkeys(value[i]))
+ chosen_label = input(str(arr_without_duplicates) + ": ")
+ for label in value[i]:
+ if label != chosen_label: # if label == chosen_label: skip it because no replacement is needed
+ nb_replacement_dep += pd.Series(self.dataset[col_name_old] == label).sum()
+ nb_replacement_arr += pd.Series(self.dataset[col_name_new] == label).sum()
+ self.dataset.loc[self.dataset[col_name_old] == label, col_name_old] = chosen_label
+ self.dataset.loc[self.dataset[col_name_new] == label, col_name_new] = chosen_label
+ size = int(self.dataset.count()[OLD_PREFIX + key]) + int(self.dataset.count()[NEW_PREFIX + key])
+ mean_error = ((nb_replacement_dep + nb_replacement_arr) / size) * 100
+ log.debug("%d IRIS have been corrected for the EV %s, corresponding to an error rate of %.0f %%", (nb_replacement_dep + nb_replacement_arr), key, mean_error)
+
+ # 3. removing outliers from data
+ count = 0
+ for outlier in self.outliers:
+ self.dataset.drop(self.dataset[self.dataset.eq(outlier).any(1)].index, inplace=True)
+ count += 1
+ log.debug("%d incorrect values removed", count)
+
+ # 4. save data
+ self.dataset.to_csv(FILE_CLEANED_DATA, index=False, encoding='utf-8')
+ log.info("Cleaned data is in %s", FILE_CLEANED_DATA)</code></pre>
+</details>
+</dd>
+<dt id="MethodCleaning.MethodCleaning.create_bar_chart"><code class="name flex">
+<span>def <span class="ident">create_bar_chart</span></span>(<span>self, name, title)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Plot before/after charts.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>name</code></strong></dt>
+<dd>a string containing the name of the EV to plot, e.g. building_type, building_usage, landscape, …</dd>
+<dt><strong><code>title</code></strong></dt>
+<dd>a string containing the title of the plot</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def create_bar_chart(self, name, title):
+ """
+ Plot before/after charts.
+
+ Args:
+ name: a string containing the name of the EV to plot, e.g. building_type, building_usage, landscape, ...
+ title: a string containing the title of the plot
+ """
+ x = np.arange(len(self.labels)) # the label locations
+ width = 0.35
+
+ fig, ax = plt.subplots()
+
+ ax.bar(x - width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ bef = ax.bar(x - width / 2, self.before, width=width, label='Avant') # before data
+ ax.bar(x + width / 2, [154 for _ in range(len(self.labels))], width=width, color="#DCDCDC") # grey bar
+ aft = ax.bar(x + width / 2, self.after, width=width, label='Après') # after data
+
+ ax.set_ylabel('Nombre de personnes')
+ plt.xticks(x, self.labels, rotation='vertical')
+ auto_label(bef, ax)
+ auto_label(aft, ax)
+ plt.tight_layout()
+ ax.legend()
+ filename = os.path.join(FOLDER_DISTRIBUTION, "distribution_" + name + ".png")
+ fig.savefig(filename)
+ ax.set_title(title)
+ plt.show()</code></pre>
+</details>
+</dd>
+<dt id="MethodCleaning.MethodCleaning.create_before_after_labels"><code class="name flex">
+<span>def <span class="ident">create_before_after_labels</span></span>(<span>self, name_dep, name_arr)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Creates the lists 'before', 'after' and 'labels' from data.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>name_dep</code></strong></dt>
+<dd>a string containing the name of the departure column, e.g. old_building_type, old_building_usage…</dd>
+<dt><strong><code>name_arr</code></strong></dt>
+<dd>a string containing the name of the arrival column, e.g. new_building_type, new_building_usage…</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def create_before_after_labels(self, name_dep, name_arr):
+ """
+ Creates the lists 'before', 'after' and 'labels' from data.
+
+ Args:
+ name_dep: a string containing the name of the departure column, e.g. old_building_type, old_building_usage...
+ name_arr: a string containing the name of the arrival column, e.g. new_building_type, new_building_usage...
+ """
+ all_repartition = {}
+ self.before = {}
+ self.after = {}
+
+ for status, value in self.dataset[name_dep].value_counts().items():
+ if name_dep == OLD_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.before:
+ self.before[status] += value
+ else:
+ self.before[status] = value
+ else:
+ self.before[status] = value # self.dataset[values_before].value_counts()[status]
+
+ for status, value in self.dataset[name_arr].value_counts().items():
+ if name_arr == NEW_PREFIX+"geographical_position": # if geo, get only the geo position (South, East, ..) and not the city
+ status = status.split(" ")[0]
+ if status in self.after:
+ self.after[status] += value
+ else:
+ self.after[status] = value
+ else:
+ self.after[status] = value # self.dataset[values_after].value_counts()[status]
+
+ # 2. merge before and after data in the same dictionary
+ for status in self.before:
+ all_repartition[status] = [self.before[status], 0]
+ for status in self.after:
+ if status not in all_repartition:
+ all_repartition[status] = [0, self.after[status]]
+ else:
+ all_repartition[status][1] = self.after[status]
+
+ # 3. convert dictionary in 3 arrays
+ self.before = []
+ self.after = []
+ self.labels = []
+ for key in all_repartition:
+ if not isinstance(key, float): # to remove nan values
+ self.before.append(all_repartition[key][0])
+ self.after.append(all_repartition[key][1])
+ self.labels.append(key)</code></pre>
+</details>
+</dd>
+<dt id="MethodCleaning.MethodCleaning.to_chart"><code class="name flex">
+<span>def <span class="ident">to_chart</span></span>(<span>self, env, name, title)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Create before/after data and plot it.
+:param env:
+:param name:
+:param title:</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>env</code></strong></dt>
+<dd>a string containing the EV to plot, e.g. building_type, building_usage, landscape…</dd>
+<dt><strong><code>name</code></strong></dt>
+<dd>a string containing the name to save the file</dd>
+<dt><strong><code>title</code></strong></dt>
+<dd>a string containing the title of the plot</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def to_chart(self, env, name, title):
+ """
+ Create before/after data and plot it.
+ :param env:
+ :param name:
+ :param title:
+ Args:
+ env: a string containing the EV to plot, e.g. building_type, building_usage, landscape...
+ name: a string containing the name to save the file
+ title: a string containing the title of the plot
+ """
+ self.create_before_after_labels(self.columns[env]["name_dep"], self.columns[env]["name_arr"])
+ self.create_bar_chart(name, title)</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="MethodCleaning.MethodCleaning" href="#MethodCleaning.MethodCleaning">MethodCleaning</a></code></h4>
+<ul class="">
+<li><code><a title="MethodCleaning.MethodCleaning.clean" href="#MethodCleaning.MethodCleaning.clean">clean</a></code></li>
+<li><code><a title="MethodCleaning.MethodCleaning.create_bar_chart" href="#MethodCleaning.MethodCleaning.create_bar_chart">create_bar_chart</a></code></li>
+<li><code><a title="MethodCleaning.MethodCleaning.create_before_after_labels" href="#MethodCleaning.MethodCleaning.create_before_after_labels">create_before_after_labels</a></code></li>
+<li><code><a title="MethodCleaning.MethodCleaning.to_chart" href="#MethodCleaning.MethodCleaning.to_chart">to_chart</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/MethodPrediction.html b/doc/MethodPrediction.html
new file mode 100644
index 00000000..eebcdaf1
--- /dev/null
+++ b/doc/MethodPrediction.html
@@ -0,0 +1,282 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>MethodPrediction API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>MethodPrediction</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+
+from area import area
+from predihood import model
+from predihood.classes.Method import Method
+from sklearn.metrics import auc, roc_curve
+from sklearn.model_selection import cross_val_score, GridSearchCV
+from sklearn.tree import DecisionTreeClassifier
+
+log = logging.getLogger(__name__)
+
+
+class MethodPrediction(Method):
+ """
+ This class represents a method for predicting EV and compute performance at a national level.
+ """
+ def __init__(self, name, dataset, classifier):
+ """
+ Constructor of the MethodPrediction class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset, classifier)
+ self.prediction = None
+ self.accuracy = 0 # in percentage
+ self.precision = 0
+ self.recall = 0
+ self.confusion_matrix = [0, 0, 0, 0]
+
+ def compute_performance(self):
+ """
+ Compute performance metrics, i.e. accuracy.
+ """
+ scores = cross_val_score(self.classifier, self.dataset.X, self.dataset.Y, cv=5)
+ self.accuracy = scores.mean() * 100
+
+ def predict(self, iris_code=None):
+ """
+ Predict one EV for the given iris. The EV to predict is stored in the dataset as "env" variable.
+ """
+ iris_object = model.get_iris_from_code(iris_code)
+ iris_area = area(model.get_coords_from_code(iris_code)) / 1000000
+ iris_population = iris_object["properties"]["raw_indicators"]["P14_POP"]
+ iris_indicators_values = []
+ iris_indicators_names = []
+ if self.dataset.normalization == "density":
+ density = iris_population/iris_area
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"] and density > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / density)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ elif self.dataset.normalization == "population":
+ for indicator in self.dataset.selected_indicators:
+ # if indicator == "P14_POP": continue # skip this indicator because of normalisation # TODO
+ if indicator in iris_object["properties"]["raw_indicators"] and iris_population > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / iris_population)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ else:
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"]:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]))
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+
+ df = pd.DataFrame([iris_indicators_values], columns=iris_indicators_names)
+ self.prediction = self.classifier.predict(df)[0]
+ log.debug("%s", self.prediction)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="MethodPrediction.MethodPrediction"><code class="flex name class">
+<span>class <span class="ident">MethodPrediction</span></span>
+<span>(</span><span>name, dataset, classifier)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>This class represents a method for predicting EV and compute performance at a national level.</p>
+<p>Constructor of the MethodPrediction class. Initialize attributes.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class MethodPrediction(Method):
+ """
+ This class represents a method for predicting EV and compute performance at a national level.
+ """
+ def __init__(self, name, dataset, classifier):
+ """
+ Constructor of the MethodPrediction class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset, classifier)
+ self.prediction = None
+ self.accuracy = 0 # in percentage
+ self.precision = 0
+ self.recall = 0
+ self.confusion_matrix = [0, 0, 0, 0]
+
+ def compute_performance(self):
+ """
+ Compute performance metrics, i.e. accuracy.
+ """
+ scores = cross_val_score(self.classifier, self.dataset.X, self.dataset.Y, cv=5)
+ self.accuracy = scores.mean() * 100
+
+ def predict(self, iris_code=None):
+ """
+ Predict one EV for the given iris. The EV to predict is stored in the dataset as "env" variable.
+ """
+ iris_object = model.get_iris_from_code(iris_code)
+ iris_area = area(model.get_coords_from_code(iris_code)) / 1000000
+ iris_population = iris_object["properties"]["raw_indicators"]["P14_POP"]
+ iris_indicators_values = []
+ iris_indicators_names = []
+ if self.dataset.normalization == "density":
+ density = iris_population/iris_area
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"] and density > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / density)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ elif self.dataset.normalization == "population":
+ for indicator in self.dataset.selected_indicators:
+ # if indicator == "P14_POP": continue # skip this indicator because of normalisation # TODO
+ if indicator in iris_object["properties"]["raw_indicators"] and iris_population > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / iris_population)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ else:
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"]:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]))
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+
+ df = pd.DataFrame([iris_indicators_values], columns=iris_indicators_names)
+ self.prediction = self.classifier.predict(df)[0]
+ log.debug("%s", self.prediction)</code></pre>
+</details>
+<h3>Ancestors</h3>
+<ul class="hlist">
+<li>predihood.classes.Method.Method</li>
+</ul>
+<h3>Methods</h3>
+<dl>
+<dt id="MethodPrediction.MethodPrediction.compute_performance"><code class="name flex">
+<span>def <span class="ident">compute_performance</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Compute performance metrics, i.e. accuracy.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def compute_performance(self):
+ """
+ Compute performance metrics, i.e. accuracy.
+ """
+ scores = cross_val_score(self.classifier, self.dataset.X, self.dataset.Y, cv=5)
+ self.accuracy = scores.mean() * 100</code></pre>
+</details>
+</dd>
+<dt id="MethodPrediction.MethodPrediction.predict"><code class="name flex">
+<span>def <span class="ident">predict</span></span>(<span>self, iris_code=None)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Predict one EV for the given iris. The EV to predict is stored in the dataset as "env" variable.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def predict(self, iris_code=None):
+ """
+ Predict one EV for the given iris. The EV to predict is stored in the dataset as "env" variable.
+ """
+ iris_object = model.get_iris_from_code(iris_code)
+ iris_area = area(model.get_coords_from_code(iris_code)) / 1000000
+ iris_population = iris_object["properties"]["raw_indicators"]["P14_POP"]
+ iris_indicators_values = []
+ iris_indicators_names = []
+ if self.dataset.normalization == "density":
+ density = iris_population/iris_area
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"] and density > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / density)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ elif self.dataset.normalization == "population":
+ for indicator in self.dataset.selected_indicators:
+ # if indicator == "P14_POP": continue # skip this indicator because of normalisation # TODO
+ if indicator in iris_object["properties"]["raw_indicators"] and iris_population > 0:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]) / iris_population)
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+ else:
+ for indicator in self.dataset.selected_indicators:
+ if indicator in iris_object["properties"]["raw_indicators"]:
+ iris_indicators_values.append(float(iris_object["properties"]["raw_indicators"][indicator]))
+ else:
+ iris_indicators_values.append(0)
+ iris_indicators_names.append(indicator)
+
+ df = pd.DataFrame([iris_indicators_values], columns=iris_indicators_names)
+ self.prediction = self.classifier.predict(df)[0]
+ log.debug("%s", self.prediction)</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="MethodPrediction.MethodPrediction" href="#MethodPrediction.MethodPrediction">MethodPrediction</a></code></h4>
+<ul class="">
+<li><code><a title="MethodPrediction.MethodPrediction.compute_performance" href="#MethodPrediction.MethodPrediction.compute_performance">compute_performance</a></code></li>
+<li><code><a title="MethodPrediction.MethodPrediction.predict" href="#MethodPrediction.MethodPrediction.predict">predict</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/MethodSelection.html b/doc/MethodSelection.html
new file mode 100644
index 00000000..3d6db8a3
--- /dev/null
+++ b/doc/MethodSelection.html
@@ -0,0 +1,272 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>MethodSelection API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>MethodSelection</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import os
+
+import matplotlib.pyplot as plt
+import numpy as np
+import seaborn as sns
+
+from predihood.classes.Method import Method
+from predihood.config import RANDOM_STATE, TITLES
+
+
+class MethodSelection(Method):
+ """
+ This class represents a method for selection a subset of indicators among all INSEE indicators.
+ """
+ def __init__(self, name, dataset, classifier=None, transform=False, parameters=None):
+ """
+ Constructor of the MethodSelection class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset, classifier)
+ self.transform = transform
+ self.threshold = None
+ self.best_indicators = None
+ self.parameters = parameters
+
+ def fit(self):
+ """
+ Fit the train data on the classifier.
+ """
+ if self.classifier is not None:
+ self.classifier.random_state = RANDOM_STATE
+ if self.transform:
+ self.classifier.fit_transform(self.dataset.X_train, self.dataset.Y_train)
+ else:
+ self.classifier.fit(self.dataset.X_train, self.dataset.Y_train)
+
+ def compute_selection(self):
+ """
+ Get results of the classifier according to its name.
+ """
+ if self.name == "feature importance ET" or self.name == "feature importance RF":
+ importance = self.classifier.feature_importances_
+ indicators_importance = {self.dataset.indicators[i]: importance[i] for i in range(len(importance))} # create a dictionary to associate each indicator with its importance
+ indicators_importance = {k: v for k, v in sorted(indicators_importance.items(), key=lambda item: item[1], reverse=True)} # order dictionary by value in descending order
+ k_best = [[key, value] for key, value in indicators_importance.items()][:self.parameters["top_k"]]
+ self.best_indicators = k_best
+ elif self.name == "heat map":
+ fig, ax = plt.subplots()
+ ax.xaxis.tick_top()
+
+ # get indicators that are fully correlated (i.e. corr=1)
+ temp_data = self.dataset.data[self.dataset.indicators][:] # get only INSEE indicators (!= CODE, AREA, EV)
+ corr_matrix = temp_data.corr(method=self.parameters["method"]).abs()
+ sns.heatmap(corr_matrix)
+ upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))
+
+ self.best_indicators = []
+ for i in range(len(upper.columns)):
+ column = upper.columns[i]
+ for k, value in upper[column].items():
+ if value == 1 and column not in self.best_indicators: # and (column, k) not in self.best_indicators and (k, column) not in self.best_indicators:
+ self.best_indicators.append(column)
+
+ if TITLES: plt.title("Correlation matrix: filtering = " + self.dataset.filtering + ", normalization = " + self.dataset.normalization)
+ plt.show()
+ else:
+ raise Exception("Unknown name. Choose among [\"feature importance ET\", \"feature importance RF\", \"heat map\"].")</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="MethodSelection.MethodSelection"><code class="flex name class">
+<span>class <span class="ident">MethodSelection</span></span>
+<span>(</span><span>name, dataset, classifier=None, transform=False, parameters=None)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>This class represents a method for selection a subset of indicators among all INSEE indicators.</p>
+<p>Constructor of the MethodSelection class. Initialize attributes.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">class MethodSelection(Method):
+ """
+ This class represents a method for selection a subset of indicators among all INSEE indicators.
+ """
+ def __init__(self, name, dataset, classifier=None, transform=False, parameters=None):
+ """
+ Constructor of the MethodSelection class. Initialize attributes.
+ """
+ Method.__init__(self, name, dataset, classifier)
+ self.transform = transform
+ self.threshold = None
+ self.best_indicators = None
+ self.parameters = parameters
+
+ def fit(self):
+ """
+ Fit the train data on the classifier.
+ """
+ if self.classifier is not None:
+ self.classifier.random_state = RANDOM_STATE
+ if self.transform:
+ self.classifier.fit_transform(self.dataset.X_train, self.dataset.Y_train)
+ else:
+ self.classifier.fit(self.dataset.X_train, self.dataset.Y_train)
+
+ def compute_selection(self):
+ """
+ Get results of the classifier according to its name.
+ """
+ if self.name == "feature importance ET" or self.name == "feature importance RF":
+ importance = self.classifier.feature_importances_
+ indicators_importance = {self.dataset.indicators[i]: importance[i] for i in range(len(importance))} # create a dictionary to associate each indicator with its importance
+ indicators_importance = {k: v for k, v in sorted(indicators_importance.items(), key=lambda item: item[1], reverse=True)} # order dictionary by value in descending order
+ k_best = [[key, value] for key, value in indicators_importance.items()][:self.parameters["top_k"]]
+ self.best_indicators = k_best
+ elif self.name == "heat map":
+ fig, ax = plt.subplots()
+ ax.xaxis.tick_top()
+
+ # get indicators that are fully correlated (i.e. corr=1)
+ temp_data = self.dataset.data[self.dataset.indicators][:] # get only INSEE indicators (!= CODE, AREA, EV)
+ corr_matrix = temp_data.corr(method=self.parameters["method"]).abs()
+ sns.heatmap(corr_matrix)
+ upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))
+
+ self.best_indicators = []
+ for i in range(len(upper.columns)):
+ column = upper.columns[i]
+ for k, value in upper[column].items():
+ if value == 1 and column not in self.best_indicators: # and (column, k) not in self.best_indicators and (k, column) not in self.best_indicators:
+ self.best_indicators.append(column)
+
+ if TITLES: plt.title("Correlation matrix: filtering = " + self.dataset.filtering + ", normalization = " + self.dataset.normalization)
+ plt.show()
+ else:
+ raise Exception("Unknown name. Choose among [\"feature importance ET\", \"feature importance RF\", \"heat map\"].")</code></pre>
+</details>
+<h3>Ancestors</h3>
+<ul class="hlist">
+<li>predihood.classes.Method.Method</li>
+</ul>
+<h3>Methods</h3>
+<dl>
+<dt id="MethodSelection.MethodSelection.compute_selection"><code class="name flex">
+<span>def <span class="ident">compute_selection</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get results of the classifier according to its name.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def compute_selection(self):
+ """
+ Get results of the classifier according to its name.
+ """
+ if self.name == "feature importance ET" or self.name == "feature importance RF":
+ importance = self.classifier.feature_importances_
+ indicators_importance = {self.dataset.indicators[i]: importance[i] for i in range(len(importance))} # create a dictionary to associate each indicator with its importance
+ indicators_importance = {k: v for k, v in sorted(indicators_importance.items(), key=lambda item: item[1], reverse=True)} # order dictionary by value in descending order
+ k_best = [[key, value] for key, value in indicators_importance.items()][:self.parameters["top_k"]]
+ self.best_indicators = k_best
+ elif self.name == "heat map":
+ fig, ax = plt.subplots()
+ ax.xaxis.tick_top()
+
+ # get indicators that are fully correlated (i.e. corr=1)
+ temp_data = self.dataset.data[self.dataset.indicators][:] # get only INSEE indicators (!= CODE, AREA, EV)
+ corr_matrix = temp_data.corr(method=self.parameters["method"]).abs()
+ sns.heatmap(corr_matrix)
+ upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))
+
+ self.best_indicators = []
+ for i in range(len(upper.columns)):
+ column = upper.columns[i]
+ for k, value in upper[column].items():
+ if value == 1 and column not in self.best_indicators: # and (column, k) not in self.best_indicators and (k, column) not in self.best_indicators:
+ self.best_indicators.append(column)
+
+ if TITLES: plt.title("Correlation matrix: filtering = " + self.dataset.filtering + ", normalization = " + self.dataset.normalization)
+ plt.show()
+ else:
+ raise Exception("Unknown name. Choose among [\"feature importance ET\", \"feature importance RF\", \"heat map\"].")</code></pre>
+</details>
+</dd>
+<dt id="MethodSelection.MethodSelection.fit"><code class="name flex">
+<span>def <span class="ident">fit</span></span>(<span>self)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Fit the train data on the classifier.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def fit(self):
+ """
+ Fit the train data on the classifier.
+ """
+ if self.classifier is not None:
+ self.classifier.random_state = RANDOM_STATE
+ if self.transform:
+ self.classifier.fit_transform(self.dataset.X_train, self.dataset.Y_train)
+ else:
+ self.classifier.fit(self.dataset.X_train, self.dataset.Y_train)</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="MethodSelection.MethodSelection" href="#MethodSelection.MethodSelection">MethodSelection</a></code></h4>
+<ul class="">
+<li><code><a title="MethodSelection.MethodSelection.compute_selection" href="#MethodSelection.MethodSelection.compute_selection">compute_selection</a></code></li>
+<li><code><a title="MethodSelection.MethodSelection.fit" href="#MethodSelection.MethodSelection.fit">fit</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/__init__.html b/doc/__init__.html
new file mode 100644
index 00000000..22bd492e
--- /dev/null
+++ b/doc/__init__.html
@@ -0,0 +1,57 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>__init__ API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>__init__</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import sys
+
+logging.basicConfig(level=logging.DEBUG)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/cleaning.html b/doc/cleaning.html
new file mode 100644
index 00000000..979cab35
--- /dev/null
+++ b/doc/cleaning.html
@@ -0,0 +1,275 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>cleaning API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>cleaning</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+
+from predihood.classes.MethodCleaning import MethodCleaning
+from predihood.config import FILE_DATA_HIL, OLD_PREFIX, NEW_PREFIX, FOLDER_DISTRIBUTION
+
+log = logging.getLogger(__name__)
+
+
+def clean():
+ """
+ Clean Home in Love data and generate some charts to study distribution of data.
+ """
+ # 1. read data from Excel file
+ data = pd.read_excel(FILE_DATA_HIL)
+
+ # 2. rename dataset's columns because of bad naming convention
+ columns = ["id", "name", "HiL_id", "sex", "age", "nb_child", "income", "monthly_charges", "tax_revenue"]
+ for elem in ["address", "country", "occupation", "rent"]: columns.append(OLD_PREFIX + elem)
+ for elem in ["address", "country", "occupation"]: columns.append(NEW_PREFIX + elem)
+ columns.append("reason")
+ for elem in ["context", "status", "building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]: columns.append(OLD_PREFIX + elem)
+ for elem in ["context", "building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]: columns.append(NEW_PREFIX + elem)
+ data.columns = columns
+ data.head()
+
+ # 3. clean data by changing misspelled values
+ cleaning = MethodCleaning("cleaning", data)
+ cleaning.clean()
+ log.info("Many plots have be generated in " + FOLDER_DISTRIBUTION)
+
+ # 4. distribution between women and men
+ women_men = data["sex"].value_counts()
+ labels = "Women", "Men"
+ number_men_women = [women_men["Femme"], women_men["Homme"]] # getting number of women and number of men
+ colors = ["salmon", "lightblue"]
+ plt.pie(number_men_women, labels=labels, colors=colors, autopct='%i%%', shadow=True, startangle=90)
+ plt.axis("equal")
+ plt.title("Distribution of gender")
+ plt.show()
+
+ # 5. distribution between ages
+ data_temp = data
+ data_temp = data_temp.dropna() # remove NaN values
+ ages_plot = []
+ total_plot = []
+ min_age, max_age = int(min(data_temp["age"])), int(max(data_temp["age"]))
+ for counter in range(min_age, max_age + 1):
+ total_plot.append(data_temp.loc[data_temp.age == float(counter), "age"].count())
+ ages_plot.append(counter)
+ mean = np.average(ages_plot, weights=total_plot)
+
+ # First view: bar chart
+ plt.bar(ages_plot, total_plot)
+ plt.axvline(x=mean, color="red") # draw median age as a line
+ plt.xlabel("Age (in years)")
+ plt.ylabel("Number of people")
+ plt.title("Distribution of age")
+ plt.show()
+
+ # Second view: histogram
+ ages = data_temp["age"]
+ plt.hist(ages, facecolor="gray", align="mid")
+ plt.xlabel("Age (in years)")
+ plt.ylabel("Number of people")
+ plt.title("Distribution of age")
+ plt.show()
+
+ # 6. distribution between incomes
+ incomes = data["income"]
+ plt.hist(incomes, facecolor="gray", align="mid")
+ plt.title("Distribution of monthly income")
+ plt.xlabel("Monthly income (in euros)")
+ plt.ylabel("Number of people")
+ plt.show()
+
+ # 7. distribution between reasons of transfer
+ transfers = data["reason"].value_counts()
+ labels = transfers.index.tolist()
+ sizes = [transfers[i] for i in range(len(transfers))]
+ plt.pie(sizes, labels=labels, autopct="%.2f", shadow=True, startangle=90)
+ plt.axis("equal")
+ plt.title("Distribution of the reason of job transfer")
+ plt.show()
+
+ # 8. distribution between geographic positions
+ geo = pd.concat([data[OLD_PREFIX + "geographical_position"], data[NEW_PREFIX + "geographical_position"]], ignore_index=True)
+ split_geo = [geo[i].split()[0] if not isinstance(geo[i], float) else "" for i in range(len(geo))]
+ set_geo = set(split_geo)
+ uniques = [split_geo.count(elem) for elem in set_geo]
+ labels = set_geo
+ plt.pie(uniques, labels=labels, autopct="%.2f", shadow=True, startangle=90)
+ plt.title("Distribution of the geographical position")
+ plt.show()
+
+ # 9. draw evolutions before and after job transfer for each EV
+ cleaning.to_chart("occupation", "status", "Evolution of status before and after job transfer")
+ cleaning.to_chart("building_type", "building_type", "Evolution of building type before and after job transfer")
+ cleaning.to_chart("building_usage", "building_usage", "Evolution of building usage before and after job transfer")
+ cleaning.to_chart("landscape", "landscapes", "Evolution of landscapes before and after job transfer")
+ cleaning.to_chart("social_class", "social", "Evolution of social classes before and after job transfer")
+ cleaning.to_chart("morphological_position", "morpho", "Evolution of morphological positions before and after job transfer")
+ cleaning.to_chart("geographical_position", "geo", "Evolution of geographical positions before and after job transfer")
+
+
+if __name__ == '__main__':
+ clean()</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="cleaning.clean"><code class="name flex">
+<span>def <span class="ident">clean</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Clean Home in Love data and generate some charts to study distribution of data.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def clean():
+ """
+ Clean Home in Love data and generate some charts to study distribution of data.
+ """
+ # 1. read data from Excel file
+ data = pd.read_excel(FILE_DATA_HIL)
+
+ # 2. rename dataset's columns because of bad naming convention
+ columns = ["id", "name", "HiL_id", "sex", "age", "nb_child", "income", "monthly_charges", "tax_revenue"]
+ for elem in ["address", "country", "occupation", "rent"]: columns.append(OLD_PREFIX + elem)
+ for elem in ["address", "country", "occupation"]: columns.append(NEW_PREFIX + elem)
+ columns.append("reason")
+ for elem in ["context", "status", "building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]: columns.append(OLD_PREFIX + elem)
+ for elem in ["context", "building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]: columns.append(NEW_PREFIX + elem)
+ data.columns = columns
+ data.head()
+
+ # 3. clean data by changing misspelled values
+ cleaning = MethodCleaning("cleaning", data)
+ cleaning.clean()
+ log.info("Many plots have be generated in " + FOLDER_DISTRIBUTION)
+
+ # 4. distribution between women and men
+ women_men = data["sex"].value_counts()
+ labels = "Women", "Men"
+ number_men_women = [women_men["Femme"], women_men["Homme"]] # getting number of women and number of men
+ colors = ["salmon", "lightblue"]
+ plt.pie(number_men_women, labels=labels, colors=colors, autopct='%i%%', shadow=True, startangle=90)
+ plt.axis("equal")
+ plt.title("Distribution of gender")
+ plt.show()
+
+ # 5. distribution between ages
+ data_temp = data
+ data_temp = data_temp.dropna() # remove NaN values
+ ages_plot = []
+ total_plot = []
+ min_age, max_age = int(min(data_temp["age"])), int(max(data_temp["age"]))
+ for counter in range(min_age, max_age + 1):
+ total_plot.append(data_temp.loc[data_temp.age == float(counter), "age"].count())
+ ages_plot.append(counter)
+ mean = np.average(ages_plot, weights=total_plot)
+
+ # First view: bar chart
+ plt.bar(ages_plot, total_plot)
+ plt.axvline(x=mean, color="red") # draw median age as a line
+ plt.xlabel("Age (in years)")
+ plt.ylabel("Number of people")
+ plt.title("Distribution of age")
+ plt.show()
+
+ # Second view: histogram
+ ages = data_temp["age"]
+ plt.hist(ages, facecolor="gray", align="mid")
+ plt.xlabel("Age (in years)")
+ plt.ylabel("Number of people")
+ plt.title("Distribution of age")
+ plt.show()
+
+ # 6. distribution between incomes
+ incomes = data["income"]
+ plt.hist(incomes, facecolor="gray", align="mid")
+ plt.title("Distribution of monthly income")
+ plt.xlabel("Monthly income (in euros)")
+ plt.ylabel("Number of people")
+ plt.show()
+
+ # 7. distribution between reasons of transfer
+ transfers = data["reason"].value_counts()
+ labels = transfers.index.tolist()
+ sizes = [transfers[i] for i in range(len(transfers))]
+ plt.pie(sizes, labels=labels, autopct="%.2f", shadow=True, startangle=90)
+ plt.axis("equal")
+ plt.title("Distribution of the reason of job transfer")
+ plt.show()
+
+ # 8. distribution between geographic positions
+ geo = pd.concat([data[OLD_PREFIX + "geographical_position"], data[NEW_PREFIX + "geographical_position"]], ignore_index=True)
+ split_geo = [geo[i].split()[0] if not isinstance(geo[i], float) else "" for i in range(len(geo))]
+ set_geo = set(split_geo)
+ uniques = [split_geo.count(elem) for elem in set_geo]
+ labels = set_geo
+ plt.pie(uniques, labels=labels, autopct="%.2f", shadow=True, startangle=90)
+ plt.title("Distribution of the geographical position")
+ plt.show()
+
+ # 9. draw evolutions before and after job transfer for each EV
+ cleaning.to_chart("occupation", "status", "Evolution of status before and after job transfer")
+ cleaning.to_chart("building_type", "building_type", "Evolution of building type before and after job transfer")
+ cleaning.to_chart("building_usage", "building_usage", "Evolution of building usage before and after job transfer")
+ cleaning.to_chart("landscape", "landscapes", "Evolution of landscapes before and after job transfer")
+ cleaning.to_chart("social_class", "social", "Evolution of social classes before and after job transfer")
+ cleaning.to_chart("morphological_position", "morpho", "Evolution of morphological positions before and after job transfer")
+ cleaning.to_chart("geographical_position", "geo", "Evolution of geographical positions before and after job transfer")</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="cleaning.clean" href="#cleaning.clean">clean</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/config.html b/doc/config.html
new file mode 100644
index 00000000..2a0bf795
--- /dev/null
+++ b/doc/config.html
@@ -0,0 +1,153 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>config API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>config</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import os
+
+from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
+from sklearn.neighbors import KNeighborsClassifier
+from sklearn.neural_network import MLPClassifier
+from sklearn.svm import SVC
+from sklearn.tree import DecisionTreeClassifier
+
+# 1. define files used by Predihood (reading and writing files)
+FOLDER_DATA = "generated_files"
+FOLDER_CLASSES = "classes"
+
+if not os.path.exists(FOLDER_DATA): os.mkdir(FOLDER_DATA)
+FOLDER_DATASETS = os.path.join(FOLDER_DATA, "datasets")
+if not os.path.exists(FOLDER_DATASETS): os.mkdir(FOLDER_DATASETS)
+FOLDER_DISTRIBUTION = os.path.join(FOLDER_DATA, "distribution-plots")
+if not os.path.exists(FOLDER_DISTRIBUTION): os.mkdir(FOLDER_DISTRIBUTION)
+FOLDER_SELECTED_INDICATORS = os.path.join(FOLDER_DATA, "selected-indicators")
+if not os.path.exists(FOLDER_SELECTED_INDICATORS): os.mkdir(FOLDER_SELECTED_INDICATORS)
+
+FILE_HIERARCHY = os.path.join(FOLDER_CLASSES, "hierarchy.csv")
+FILE_DATA_HIL = os.path.join(FOLDER_CLASSES, "data.xlsx")
+FILE_GROUPING = os.path.join(FOLDER_CLASSES, "regrouping.csv")
+FILE_CLEANED_DATA = os.path.join(FOLDER_DATA, "cleaned_data.csv")
+FILE_ENV = os.path.join(FOLDER_DATASETS, "data.csv")
+FILE_SIMILARITIES = os.path.join(FOLDER_DATA, "similarities.csv")
+FILE_LIST_DISTRIBUTION = os.path.join(FOLDER_SELECTED_INDICATORS, "selection-distribution.csv")
+FILE_MANUAL_ASSESSMENT = os.path.join(FOLDER_CLASSES, "manual_assessment.csv")
+
+# 2. define some constants
+TITLES = False # set True to display titles on plots
+OLD_PREFIX, NEW_PREFIX = "old_", "new_"
+RANDOM_STATE = 0 # make classifiers deterministic
+TRAIN_SIZE, TEST_SIZE = 0.8, 0.2
+TOPS_K = [10, 20, 30, 40, 50, 75, 100] # define top-k to generate lists of selected indicators
+
+# 3. define available classifiers for the algorithmic interface
+AVAILABLE_CLASSIFIERS = {
+ "RandomForestClassifier": RandomForestClassifier,
+ "KNeighborsClassifier": KNeighborsClassifier,
+ "DecisionTreeClassifier": DecisionTreeClassifier,
+ "SVC": SVC,
+ "AdaBoostClassifier": AdaBoostClassifier,
+ "MLPClassifier": MLPClassifier
+}
+
+# 4. define EV, the possible values for each one and their translation (because data come from a French company).
+ENVIRONMENT_VALUES = {
+ "building_type": {
+ "Maisons": "Houses",
+ "Mixte": "Mixed",
+ "Immeubles": "Towers",
+ "Grand ensemble": "Housing estates",
+ "Lotissement": "Housing subdivisions"
+ },
+ "building_usage": {
+ "Résidentiel": "Housing",
+ "Commerçant": "Shopping",
+ "Autres activités": "Other activities",
+ },
+ "landscape": {
+ "Urbanisé": "Urban",
+ "Espaces verts": "Green areas",
+ "Arboré": "Forest",
+ "Agricole": "Countryside",
+ },
+ "morphological_position": {
+ "Central": "Central",
+ "Urbain": "Urban",
+ "Péri-urbain": "Peri-urban",
+ "Rural": "Rural"
+ },
+ "geographical_position": {
+ "Centre": "Centre",
+ "Nord": "North",
+ "Sud": "South",
+ "Est": "East",
+ "Ouest": "West",
+ "Nord-Est": "North East",
+ "Sud-Est": "South East",
+ "Nord-Ouest": "North West",
+ "Sud-Ouest": "South West"
+ },
+ "social_class": {
+ "Popu": "Lower",
+ "Moyen-inf": "Lower middle",
+ "Moyen": "Middle",
+ "Moyen-sup": "Upper middle",
+ "Sup": "Upper"
+ }
+}
+
+# names of EV, i.e. ['building_type', 'building_usage', 'landscape', 'morphological_position', 'geographical_position', 'social_class']
+ENVIRONMENT_VARIABLES = list(ENVIRONMENT_VALUES.keys())
+
+# translation of each value from French to English, ie. without the level with EV' names
+TRANSLATION = {}
+for ev in ENVIRONMENT_VALUES:
+ for key in ENVIRONMENT_VALUES[ev]:
+ TRANSLATION[key] = ENVIRONMENT_VALUES[ev][key]</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/main.html b/doc/main.html
new file mode 100644
index 00000000..cd041ff0
--- /dev/null
+++ b/doc/main.html
@@ -0,0 +1,683 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>main API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>main</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">#!/usr/bin/env python
+# encoding: utf-8
+# =============================================================================
+# main.py: runs the Flask server (http://flask.pocoo.org/) using routes
+# =============================================================================
+
+import json
+import logging
+import os
+import webbrowser
+
+from flask import Flask, flash, render_template, request, send_from_directory
+from predihood import model
+from predihood.classes.Data import Data
+from predihood.config import AVAILABLE_CLASSIFIERS, TOPS_K, ENVIRONMENT_VALUES
+from predihood.predict import compute_all_accuracies, predict_one_iris
+from predihood.utility_functions import signature, get_classifier, set_classifier, add_assessment_to_file
+from sklearn.utils._testing import ignore_warnings
+
+log = logging.getLogger(__name__)
+
+app = Flask(__name__)
+app.config["JSON_SORT_KEYS"] = False
+url = "http://127.0.0.1:8081/"
+
+
+@app.route('/', defaults={'page': None})
+def index(page):
+ """
+ Render the main page of the interface, i.e. `index.html`.
+
+ Args:
+ page: The name of the specific page to display.
+
+ Returns:
+ The page to display.
+ """
+ if not model.db.connection_status: # if no connection, display a flashing message
+ flash("Could not connect to the MongoDB database ! Check the connection.", "danger")
+ return render_template('index.html')
+
+
+@app.route('/algorithms.html', methods=["GET"])
+def get_algorithms_page():
+ """
+ Render the page of the algorithmic interface, i.e. `algorithms.html`.
+
+ Returns:
+ The page to display.
+ """
+ return render_template('algorithms.html')
+
+
+@app.route('/details-iris.html', methods=["GET"])
+def get_details_iris():
+ """
+ Get all information about the given IRIS.
+
+ Returns:
+ A page that contains all information about the given IRIS (descriptive, grouped and raw indicators).
+ """
+ code_iris = request.args['code_iris']
+ iris = model.get_iris_from_code(code_iris)
+ dict_code_label = model.parse_json_to_dict(model.json_iris_indicator_code_to_label)
+ if iris is None:
+ flash("No corresponding iris for code " + code_iris + ".", "warning")
+ return render_template('details-iris.html', iris=iris, dict_code_label=dict_code_label)
+
+
+@app.route('/getClassifiers', methods=["GET"])
+def get_classifiers():
+ """
+ Get list of available classifiers (stored in AVAILABLE_CLASSIFIERS).
+
+ Returns:
+ A list containing the names of the available classifiers.
+ """
+ return {"classifiers": list(AVAILABLE_CLASSIFIERS.keys())}
+
+
+@app.route('/getParameters', methods=["GET"])
+def get_parameters():
+ """
+ Get parameters of the given classifier by give its name.
+
+ Returns:
+ A dictionary containing for each parameter its name, its types, its default value, and its definition.
+ """
+ if 'name' in request.args:
+ name = request.args['name']
+ parameters = signature(name)
+ return parameters
+ else:
+ return None
+
+
+@app.route('/run', methods=["GET"])
+@ignore_warnings(category=FutureWarning)
+def run_algorithm():
+ """
+ Run classifier on data with the specified parameters.
+
+ Returns:
+ The computed accuracies for each EV and each list. The top-k are also returned.
+ """
+ # 1. get parameters specified by the user
+ clf_name = request.args['clf']
+ parameters = json.loads(request.args['parameters'])
+ train_size = parameters["train_size"]
+ test_size = parameters["test_size"]
+ remove_outliers = parameters["remove_outliers"]
+ remove_rural = parameters["remove_rural"]
+
+ # 2. create an instance of the given classifier and tune it with user's parameters
+ clf = get_classifier(clf_name)
+ clf = set_classifier(clf, parameters)
+ log.info(clf)
+
+ # 3. run experiment on data to get accuracies for each EV and each list of selected indicators
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ accuracies = compute_all_accuracies(data, clf, train_size, test_size, remove_outliers, remove_rural)
+ return {"results": accuracies, "tops_k": TOPS_K}
+
+
+@app.route('/predict_iris', methods=["GET"])
+def predict_iris():
+ """
+ Predict the environment (i.e. 6 EV) of the given IRIS.
+
+ Returns:
+ Predictions for each EV
+ """
+ iris_code_to_predict = request.args['iris_code']
+ clf_name = request.args['algorithm_name']
+ clf = get_classifier(clf_name)
+
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ predictions = predict_one_iris(iris_code_to_predict, data, clf, 0.8, 0.2, False) # clf
+ return {"predictions": predictions}
+
+
+@app.route('/getIrisPolygon', methods=["GET"])
+def get_iris_for_polygon():
+ """
+ Get the list of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ A list of the IRIS that are in the given polygon.
+ """
+ lat1 = float(request.args['lat1'])
+ lng1 = float(request.args['lng1'])
+ lat2 = float(request.args['lat2'])
+ lng2 = float(request.args['lng2'])
+ iris = model.get_iris_for_polygon(lat1, lng1, lat2, lng2)
+ if iris is None or len(iris) == 0:
+ flash("No iris found in the area.", "warning")
+ else:
+ flash(str(len(iris)) + " iris found in the area.", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})
+
+
+@app.route('/countIrisPolygon', methods=["GET"])
+def count_iris_for_polygon():
+ """
+ Count the number of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ The number of IRIS that are in the given polygon.
+ """
+ lat1 = float(request.args['lat1'])
+ lng1 = float(request.args['lng1'])
+ lat2 = float(request.args['lat2'])
+ lng2 = float(request.args['lng2'])
+ nb_iris = model.count_iris_for_polygon(lat1, lng1, lat2, lng2)
+ return json.dumps({'status': 'OK', 'nbIris': nb_iris})
+
+
+@app.route('/searchCode', methods=["GET"])
+def get_iris_from_code():
+ """
+ Get an IRIS object (represented by a dictionary) given its code. The code is a string of 9 digits.
+
+ Returns:
+ An object that represents the IRIS corresponding to the given code.
+ """
+ code_iris = request.args['codeIris']
+ iris = model.get_iris_from_code(code_iris)
+ if iris is None:
+ flash("No corresponding iris for code " + code_iris + ".", "warning")
+ else:
+ flash("Found iris " + code_iris + ".", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})
+
+
+@app.route('/searchName', methods=["GET"])
+def get_iris_from_name():
+ """
+ Get IRIS given its name. The search is done on IRIS's name and IRIS' city.
+
+ Returns:
+ A list of IRIS corresponding to the given name.
+ """
+ query = request.args['querySearch']
+ iris = model.get_iris_from_name(query)
+ if iris is None or len(iris) == 0:
+ flash("No corresponding iris for query " + query + ".", "warning")
+ else:
+ flash(str(len(iris)) + " iris found for query " + query + ".", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})
+
+
+@app.route('/add_iris_to_csv', methods=["GET"])
+def add_iris_to_csv():
+ """
+ Adds an assessed IRIS to a CSV file. Not available in production mode.
+
+ Returns:
+ A message that explain the status of the request, i.e. OK if the IRIS has been added, KO else.
+ """
+ assessed_values = []
+ for env in ENVIRONMENT_VALUES:
+ assessed_values.append(request.args[env])
+ message = add_assessment_to_file(request.args['code_iris'], assessed_values)
+ return json.dumps({"status": message})
+
+
+@app.route('/favicon.ico')
+@app.route('/<page>/favicon.ico')
+def favicon():
+ """
+ Display the favicon.
+
+ Returns:
+ The favicon.
+ """
+ return send_from_directory(os.path.join(app.root_path, 'static'), 'favicon.png', mimetype='image/favicon.png')
+
+
+if __name__ == '__main__':
+ webbrowser.open_new(url)
+ app.config['SEND_FILE_MAX_AGE_DEFAULT'] = 0 # do not cache files, especially static files such as JS
+ app.secret_key = 's3k_5Et#fL45k_#ranD0m-(StuF7)'
+ app.run(port=8081) # debug = True</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="main.add_iris_to_csv"><code class="name flex">
+<span>def <span class="ident">add_iris_to_csv</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Adds an assessed IRIS to a CSV file. Not available in production mode.</p>
+<h2 id="returns">Returns</h2>
+<p>A message that explain the status of the request, i.e. OK if the IRIS has been added, KO else.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/add_iris_to_csv', methods=["GET"])
+def add_iris_to_csv():
+ """
+ Adds an assessed IRIS to a CSV file. Not available in production mode.
+
+ Returns:
+ A message that explain the status of the request, i.e. OK if the IRIS has been added, KO else.
+ """
+ assessed_values = []
+ for env in ENVIRONMENT_VALUES:
+ assessed_values.append(request.args[env])
+ message = add_assessment_to_file(request.args['code_iris'], assessed_values)
+ return json.dumps({"status": message})</code></pre>
+</details>
+</dd>
+<dt id="main.count_iris_for_polygon"><code class="name flex">
+<span>def <span class="ident">count_iris_for_polygon</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Count the number of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.</p>
+<h2 id="returns">Returns</h2>
+<p>The number of IRIS that are in the given polygon.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/countIrisPolygon', methods=["GET"])
+def count_iris_for_polygon():
+ """
+ Count the number of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ The number of IRIS that are in the given polygon.
+ """
+ lat1 = float(request.args['lat1'])
+ lng1 = float(request.args['lng1'])
+ lat2 = float(request.args['lat2'])
+ lng2 = float(request.args['lng2'])
+ nb_iris = model.count_iris_for_polygon(lat1, lng1, lat2, lng2)
+ return json.dumps({'status': 'OK', 'nbIris': nb_iris})</code></pre>
+</details>
+</dd>
+<dt id="main.favicon"><code class="name flex">
+<span>def <span class="ident">favicon</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Display the favicon.</p>
+<h2 id="returns">Returns</h2>
+<p>The favicon.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/favicon.ico')
+@app.route('/<page>/favicon.ico')
+def favicon():
+ """
+ Display the favicon.
+
+ Returns:
+ The favicon.
+ """
+ return send_from_directory(os.path.join(app.root_path, 'static'), 'favicon.png', mimetype='image/favicon.png')</code></pre>
+</details>
+</dd>
+<dt id="main.get_algorithms_page"><code class="name flex">
+<span>def <span class="ident">get_algorithms_page</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Render the page of the algorithmic interface, i.e. <code>algorithms.html</code>.</p>
+<h2 id="returns">Returns</h2>
+<p>The page to display.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/algorithms.html', methods=["GET"])
+def get_algorithms_page():
+ """
+ Render the page of the algorithmic interface, i.e. `algorithms.html`.
+
+ Returns:
+ The page to display.
+ """
+ return render_template('algorithms.html')</code></pre>
+</details>
+</dd>
+<dt id="main.get_classifiers"><code class="name flex">
+<span>def <span class="ident">get_classifiers</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get list of available classifiers (stored in AVAILABLE_CLASSIFIERS).</p>
+<h2 id="returns">Returns</h2>
+<p>A list containing the names of the available classifiers.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/getClassifiers', methods=["GET"])
+def get_classifiers():
+ """
+ Get list of available classifiers (stored in AVAILABLE_CLASSIFIERS).
+
+ Returns:
+ A list containing the names of the available classifiers.
+ """
+ return {"classifiers": list(AVAILABLE_CLASSIFIERS.keys())}</code></pre>
+</details>
+</dd>
+<dt id="main.get_details_iris"><code class="name flex">
+<span>def <span class="ident">get_details_iris</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get all information about the given IRIS.</p>
+<h2 id="returns">Returns</h2>
+<p>A page that contains all information about the given IRIS (descriptive, grouped and raw indicators).</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/details-iris.html', methods=["GET"])
+def get_details_iris():
+ """
+ Get all information about the given IRIS.
+
+ Returns:
+ A page that contains all information about the given IRIS (descriptive, grouped and raw indicators).
+ """
+ code_iris = request.args['code_iris']
+ iris = model.get_iris_from_code(code_iris)
+ dict_code_label = model.parse_json_to_dict(model.json_iris_indicator_code_to_label)
+ if iris is None:
+ flash("No corresponding iris for code " + code_iris + ".", "warning")
+ return render_template('details-iris.html', iris=iris, dict_code_label=dict_code_label)</code></pre>
+</details>
+</dd>
+<dt id="main.get_iris_for_polygon"><code class="name flex">
+<span>def <span class="ident">get_iris_for_polygon</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get the list of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.</p>
+<h2 id="returns">Returns</h2>
+<p>A list of the IRIS that are in the given polygon.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/getIrisPolygon', methods=["GET"])
+def get_iris_for_polygon():
+ """
+ Get the list of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ A list of the IRIS that are in the given polygon.
+ """
+ lat1 = float(request.args['lat1'])
+ lng1 = float(request.args['lng1'])
+ lat2 = float(request.args['lat2'])
+ lng2 = float(request.args['lng2'])
+ iris = model.get_iris_for_polygon(lat1, lng1, lat2, lng2)
+ if iris is None or len(iris) == 0:
+ flash("No iris found in the area.", "warning")
+ else:
+ flash(str(len(iris)) + " iris found in the area.", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})</code></pre>
+</details>
+</dd>
+<dt id="main.get_iris_from_code"><code class="name flex">
+<span>def <span class="ident">get_iris_from_code</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get an IRIS object (represented by a dictionary) given its code. The code is a string of 9 digits.</p>
+<h2 id="returns">Returns</h2>
+<p>An object that represents the IRIS corresponding to the given code.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/searchCode', methods=["GET"])
+def get_iris_from_code():
+ """
+ Get an IRIS object (represented by a dictionary) given its code. The code is a string of 9 digits.
+
+ Returns:
+ An object that represents the IRIS corresponding to the given code.
+ """
+ code_iris = request.args['codeIris']
+ iris = model.get_iris_from_code(code_iris)
+ if iris is None:
+ flash("No corresponding iris for code " + code_iris + ".", "warning")
+ else:
+ flash("Found iris " + code_iris + ".", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})</code></pre>
+</details>
+</dd>
+<dt id="main.get_iris_from_name"><code class="name flex">
+<span>def <span class="ident">get_iris_from_name</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get IRIS given its name. The search is done on IRIS's name and IRIS' city.</p>
+<h2 id="returns">Returns</h2>
+<p>A list of IRIS corresponding to the given name.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/searchName', methods=["GET"])
+def get_iris_from_name():
+ """
+ Get IRIS given its name. The search is done on IRIS's name and IRIS' city.
+
+ Returns:
+ A list of IRIS corresponding to the given name.
+ """
+ query = request.args['querySearch']
+ iris = model.get_iris_from_name(query)
+ if iris is None or len(iris) == 0:
+ flash("No corresponding iris for query " + query + ".", "warning")
+ else:
+ flash(str(len(iris)) + " iris found for query " + query + ".", "success")
+ return json.dumps({'status': 'OK', 'geojson': iris})</code></pre>
+</details>
+</dd>
+<dt id="main.get_parameters"><code class="name flex">
+<span>def <span class="ident">get_parameters</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get parameters of the given classifier by give its name.</p>
+<h2 id="returns">Returns</h2>
+<p>A dictionary containing for each parameter its name, its types, its default value, and its definition.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/getParameters', methods=["GET"])
+def get_parameters():
+ """
+ Get parameters of the given classifier by give its name.
+
+ Returns:
+ A dictionary containing for each parameter its name, its types, its default value, and its definition.
+ """
+ if 'name' in request.args:
+ name = request.args['name']
+ parameters = signature(name)
+ return parameters
+ else:
+ return None</code></pre>
+</details>
+</dd>
+<dt id="main.index"><code class="name flex">
+<span>def <span class="ident">index</span></span>(<span>page)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Render the main page of the interface, i.e. <code>index.html</code>.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>page</code></strong></dt>
+<dd>The name of the specific page to display.</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>The page to display.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/', defaults={'page': None})
+def index(page):
+ """
+ Render the main page of the interface, i.e. `index.html`.
+
+ Args:
+ page: The name of the specific page to display.
+
+ Returns:
+ The page to display.
+ """
+ if not model.db.connection_status: # if no connection, display a flashing message
+ flash("Could not connect to the MongoDB database ! Check the connection.", "danger")
+ return render_template('index.html')</code></pre>
+</details>
+</dd>
+<dt id="main.predict_iris"><code class="name flex">
+<span>def <span class="ident">predict_iris</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Predict the environment (i.e. 6 EV) of the given IRIS.</p>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>Predictions for each EV</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/predict_iris', methods=["GET"])
+def predict_iris():
+ """
+ Predict the environment (i.e. 6 EV) of the given IRIS.
+
+ Returns:
+ Predictions for each EV
+ """
+ iris_code_to_predict = request.args['iris_code']
+ clf_name = request.args['algorithm_name']
+ clf = get_classifier(clf_name)
+
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ predictions = predict_one_iris(iris_code_to_predict, data, clf, 0.8, 0.2, False) # clf
+ return {"predictions": predictions}</code></pre>
+</details>
+</dd>
+<dt id="main.run_algorithm"><code class="name flex">
+<span>def <span class="ident">run_algorithm</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Run classifier on data with the specified parameters.</p>
+<h2 id="returns">Returns</h2>
+<p>The computed accuracies for each EV and each list. The top-k are also returned.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@app.route('/run', methods=["GET"])
+@ignore_warnings(category=FutureWarning)
+def run_algorithm():
+ """
+ Run classifier on data with the specified parameters.
+
+ Returns:
+ The computed accuracies for each EV and each list. The top-k are also returned.
+ """
+ # 1. get parameters specified by the user
+ clf_name = request.args['clf']
+ parameters = json.loads(request.args['parameters'])
+ train_size = parameters["train_size"]
+ test_size = parameters["test_size"]
+ remove_outliers = parameters["remove_outliers"]
+ remove_rural = parameters["remove_rural"]
+
+ # 2. create an instance of the given classifier and tune it with user's parameters
+ clf = get_classifier(clf_name)
+ clf = set_classifier(clf, parameters)
+ log.info(clf)
+
+ # 3. run experiment on data to get accuracies for each EV and each list of selected indicators
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ accuracies = compute_all_accuracies(data, clf, train_size, test_size, remove_outliers, remove_rural)
+ return {"results": accuracies, "tops_k": TOPS_K}</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="main.add_iris_to_csv" href="#main.add_iris_to_csv">add_iris_to_csv</a></code></li>
+<li><code><a title="main.count_iris_for_polygon" href="#main.count_iris_for_polygon">count_iris_for_polygon</a></code></li>
+<li><code><a title="main.favicon" href="#main.favicon">favicon</a></code></li>
+<li><code><a title="main.get_algorithms_page" href="#main.get_algorithms_page">get_algorithms_page</a></code></li>
+<li><code><a title="main.get_classifiers" href="#main.get_classifiers">get_classifiers</a></code></li>
+<li><code><a title="main.get_details_iris" href="#main.get_details_iris">get_details_iris</a></code></li>
+<li><code><a title="main.get_iris_for_polygon" href="#main.get_iris_for_polygon">get_iris_for_polygon</a></code></li>
+<li><code><a title="main.get_iris_from_code" href="#main.get_iris_from_code">get_iris_from_code</a></code></li>
+<li><code><a title="main.get_iris_from_name" href="#main.get_iris_from_name">get_iris_from_name</a></code></li>
+<li><code><a title="main.get_parameters" href="#main.get_parameters">get_parameters</a></code></li>
+<li><code><a title="main.index" href="#main.index">index</a></code></li>
+<li><code><a title="main.predict_iris" href="#main.predict_iris">predict_iris</a></code></li>
+<li><code><a title="main.run_algorithm" href="#main.run_algorithm">run_algorithm</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/model.html b/doc/model.html
new file mode 100644
index 00000000..b976e00f
--- /dev/null
+++ b/doc/model.html
@@ -0,0 +1,490 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>model API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>model</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">#!/usr/bin/env python
+# encoding: utf-8
+# =============================================================================
+# model.py: methods for getting data from MongoDB (using mongiris) and transforming it
+# =============================================================================
+
+import json
+import os
+import re
+
+from mongiris.api import Mongiris
+
+# connection to the IRIS collection in MongoDB
+db = Mongiris()
+iris_collection = db.collection_iris
+json_iris_indicator_code_to_label = 'static/data/dictionnaire-indicateurs.json'
+
+
+def get_iris_for_polygon(lat1, lng1, lat2, lng2):
+ """
+ Get IRIS that are in a box represented by given coordinates.
+
+ Args:
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
+ Returns:
+ a list of IRIS that are in the box defined by given latitudes and longitudes
+ """
+ # polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ polygon = Mongiris.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ # iris = db.geo_within(iris_collection, polygon)
+ iris = db.intersect(iris_collection, polygon)
+ return iris
+
+
+def count_iris_for_polygon(lat1, lng1, lat2, lng2):
+ """
+ Count IRIS that are in a box represented by given coordinates.
+
+ Args:
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
+ Returns:
+ an integer representing the number of IRIS that are in the box defined by given latitudes and longitudes
+ """
+ polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ iris = db.geo_within(iris_collection, polygon)
+ if iris is None: return 0
+ return len(iris)
+
+
+def get_iris_from_code(code_iris):
+ """
+ Get an IRIS given its code (9 digits).
+
+ Args:
+ code_iris: a string corresponding to the code of the IRIS (the code should be a 9 digits string)
+
+ Returns:
+ an object that represents the IRIS corresponding to the given code
+ """
+ iris = db.get_iris_from_code(code_iris)
+ return iris
+
+
+def get_iris_from_name(name):
+ """
+ Get an IRIS given its name.
+
+ Args:
+ name: a string corresponding to the name of the IRIS
+
+ Returns:
+ an object that represents the IRIS corresponding to the given name
+ """
+ # the query string (name) is searched in both the name of the iris and the name of the city
+ regx = re.compile(name, re.IGNORECASE)
+ query_clause = {"$or": [{"properties.NOM_IRIS": {"$regex": regx}}, {"properties.NOM_COM": {"$regex": regx}}]}
+ iris = db.find_documents(iris_collection, query_clause)
+ return iris
+
+
+def parse_json_to_dict(json_file_path):
+ """
+ Convert a JSON file to a dictionary object.
+
+ Args:
+ json_file_path: a string containing the path of the file to load
+
+ Returns:
+ a dictionary containing the data in the file located in the given path
+ """
+ assert(os.path.isfile(json_file_path))
+ with open(json_file_path) as data_file:
+ data = json.load(data_file)
+ data_file.close()
+ return data
+
+
+def get_coords_from_code(code):
+ """
+ Get geometry of the IRIS corresponding to the given code.
+
+ Args:
+ code: a string corresponding to the code of the IRIS. The code should be a 9 digits string
+
+ Returns:
+ a list representing coordinates of the IRIS
+ """
+ iris = db.get_iris_from_code(code)
+ if iris:
+ return db.get_geojson_polygon(iris["geometry"]["coordinates"])
+ else:
+ return None
+
+
+def get_indicators_list():
+ """
+ Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
+ Returns:
+ a list containing all names of indicators, e.g. ['POP0002', 'POP0204', ..., 'P14_RP_MAISON']
+ """
+ list_indicators = db.find_all(db.collection_indic)
+ return [indicator["short_label"] for indicator in list_indicators]
+
+
+def get_indicators_dict():
+ """
+ Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
+ Returns:
+ a dictionary containing all names of indicators, e.g. {'POP0002': 'Population aged from 0 to 2 y.o.', ..., 'P14_RP_MAISON': 'Number of principal residences'}
+ """
+ list_indicators = db.find_all(db.collection_indic)
+ return {indicator["short_label"]: indicator["full_label"] for indicator in list_indicators}</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="model.count_iris_for_polygon"><code class="name flex">
+<span>def <span class="ident">count_iris_for_polygon</span></span>(<span>lat1, lng1, lat2, lng2)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Count IRIS that are in a box represented by given coordinates.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>lat1</code></strong></dt>
+<dd>a float for latitude of the first point of the box</dd>
+<dt><strong><code>lng1</code></strong></dt>
+<dd>a float for longitude of the first point of the box</dd>
+<dt><strong><code>lat2</code></strong></dt>
+<dd>a float for latitude of the second point of the box</dd>
+<dt><strong><code>lng2</code></strong></dt>
+<dd>a float for longitude of the second point of the box</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>an integer representing the number</code> of <code>IRIS that are in the box defined by given latitudes and longitudes</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def count_iris_for_polygon(lat1, lng1, lat2, lng2):
+ """
+ Count IRIS that are in a box represented by given coordinates.
+
+ Args:
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
+ Returns:
+ an integer representing the number of IRIS that are in the box defined by given latitudes and longitudes
+ """
+ polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ iris = db.geo_within(iris_collection, polygon)
+ if iris is None: return 0
+ return len(iris)</code></pre>
+</details>
+</dd>
+<dt id="model.get_coords_from_code"><code class="name flex">
+<span>def <span class="ident">get_coords_from_code</span></span>(<span>code)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get geometry of the IRIS corresponding to the given code.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>code</code></strong></dt>
+<dd>a string corresponding to the code of the IRIS. The code should be a 9 digits string</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a list representing coordinates</code> of <code>the IRIS</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_coords_from_code(code):
+ """
+ Get geometry of the IRIS corresponding to the given code.
+
+ Args:
+ code: a string corresponding to the code of the IRIS. The code should be a 9 digits string
+
+ Returns:
+ a list representing coordinates of the IRIS
+ """
+ iris = db.get_iris_from_code(code)
+ if iris:
+ return db.get_geojson_polygon(iris["geometry"]["coordinates"])
+ else:
+ return None</code></pre>
+</details>
+</dd>
+<dt id="model.get_indicators_dict"><code class="name flex">
+<span>def <span class="ident">get_indicators_dict</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.</p>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a dictionary containing all names</code> of <code>indicators, e.g. {'POP0002': 'Population aged from 0 to 2 y.o.', ..., 'P14_RP_MAISON': 'Number</code> of <code>principal residences'}</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_indicators_dict():
+ """
+ Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
+ Returns:
+ a dictionary containing all names of indicators, e.g. {'POP0002': 'Population aged from 0 to 2 y.o.', ..., 'P14_RP_MAISON': 'Number of principal residences'}
+ """
+ list_indicators = db.find_all(db.collection_indic)
+ return {indicator["short_label"]: indicator["full_label"] for indicator in list_indicators}</code></pre>
+</details>
+</dd>
+<dt id="model.get_indicators_list"><code class="name flex">
+<span>def <span class="ident">get_indicators_list</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.</p>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a list containing all names</code> of <code>indicators, e.g. ['POP0002', 'POP0204', ..., 'P14_RP_MAISON']</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_indicators_list():
+ """
+ Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
+ Returns:
+ a list containing all names of indicators, e.g. ['POP0002', 'POP0204', ..., 'P14_RP_MAISON']
+ """
+ list_indicators = db.find_all(db.collection_indic)
+ return [indicator["short_label"] for indicator in list_indicators]</code></pre>
+</details>
+</dd>
+<dt id="model.get_iris_for_polygon"><code class="name flex">
+<span>def <span class="ident">get_iris_for_polygon</span></span>(<span>lat1, lng1, lat2, lng2)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get IRIS that are in a box represented by given coordinates.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>lat1</code></strong></dt>
+<dd>a float for latitude of the first point of the box</dd>
+<dt><strong><code>lng1</code></strong></dt>
+<dd>a float for longitude of the first point of the box</dd>
+<dt><strong><code>lat2</code></strong></dt>
+<dd>a float for latitude of the second point of the box</dd>
+<dt><strong><code>lng2</code></strong></dt>
+<dd>a float for longitude of the second point of the box</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a list</code> of <code>IRIS that are in the box defined by given latitudes and longitudes</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_iris_for_polygon(lat1, lng1, lat2, lng2):
+ """
+ Get IRIS that are in a box represented by given coordinates.
+
+ Args:
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
+ Returns:
+ a list of IRIS that are in the box defined by given latitudes and longitudes
+ """
+ # polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ polygon = Mongiris.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
+ # iris = db.geo_within(iris_collection, polygon)
+ iris = db.intersect(iris_collection, polygon)
+ return iris</code></pre>
+</details>
+</dd>
+<dt id="model.get_iris_from_code"><code class="name flex">
+<span>def <span class="ident">get_iris_from_code</span></span>(<span>code_iris)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get an IRIS given its code (9 digits).</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>code_iris</code></strong></dt>
+<dd>a string corresponding to the code of the IRIS (the code should be a 9 digits string)</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>an object that represents the IRIS corresponding to the given code</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_iris_from_code(code_iris):
+ """
+ Get an IRIS given its code (9 digits).
+
+ Args:
+ code_iris: a string corresponding to the code of the IRIS (the code should be a 9 digits string)
+
+ Returns:
+ an object that represents the IRIS corresponding to the given code
+ """
+ iris = db.get_iris_from_code(code_iris)
+ return iris</code></pre>
+</details>
+</dd>
+<dt id="model.get_iris_from_name"><code class="name flex">
+<span>def <span class="ident">get_iris_from_name</span></span>(<span>name)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get an IRIS given its name.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>name</code></strong></dt>
+<dd>a string corresponding to the name of the IRIS</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>an object that represents the IRIS corresponding to the given name</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_iris_from_name(name):
+ """
+ Get an IRIS given its name.
+
+ Args:
+ name: a string corresponding to the name of the IRIS
+
+ Returns:
+ an object that represents the IRIS corresponding to the given name
+ """
+ # the query string (name) is searched in both the name of the iris and the name of the city
+ regx = re.compile(name, re.IGNORECASE)
+ query_clause = {"$or": [{"properties.NOM_IRIS": {"$regex": regx}}, {"properties.NOM_COM": {"$regex": regx}}]}
+ iris = db.find_documents(iris_collection, query_clause)
+ return iris</code></pre>
+</details>
+</dd>
+<dt id="model.parse_json_to_dict"><code class="name flex">
+<span>def <span class="ident">parse_json_to_dict</span></span>(<span>json_file_path)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Convert a JSON file to a dictionary object.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>json_file_path</code></strong></dt>
+<dd>a string containing the path of the file to load</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a dictionary containing the data in the file located in the given path</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def parse_json_to_dict(json_file_path):
+ """
+ Convert a JSON file to a dictionary object.
+
+ Args:
+ json_file_path: a string containing the path of the file to load
+
+ Returns:
+ a dictionary containing the data in the file located in the given path
+ """
+ assert(os.path.isfile(json_file_path))
+ with open(json_file_path) as data_file:
+ data = json.load(data_file)
+ data_file.close()
+ return data</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="model.count_iris_for_polygon" href="#model.count_iris_for_polygon">count_iris_for_polygon</a></code></li>
+<li><code><a title="model.get_coords_from_code" href="#model.get_coords_from_code">get_coords_from_code</a></code></li>
+<li><code><a title="model.get_indicators_dict" href="#model.get_indicators_dict">get_indicators_dict</a></code></li>
+<li><code><a title="model.get_indicators_list" href="#model.get_indicators_list">get_indicators_list</a></code></li>
+<li><code><a title="model.get_iris_for_polygon" href="#model.get_iris_for_polygon">get_iris_for_polygon</a></code></li>
+<li><code><a title="model.get_iris_from_code" href="#model.get_iris_from_code">get_iris_from_code</a></code></li>
+<li><code><a title="model.get_iris_from_name" href="#model.get_iris_from_name">get_iris_from_name</a></code></li>
+<li><code><a title="model.parse_json_to_dict" href="#model.parse_json_to_dict">parse_json_to_dict</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/predict.html b/doc/predict.html
new file mode 100644
index 00000000..78813fa5
--- /dev/null
+++ b/doc/predict.html
@@ -0,0 +1,345 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>predict API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>predict</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import logging
+from collections import OrderedDict
+
+from predihood.classes.Data import Data
+from predihood.classes.Dataset import Dataset
+from predihood.classes.MethodPrediction import MethodPrediction
+from predihood.config import ENVIRONMENT_VARIABLES, TRAIN_SIZE, TEST_SIZE
+from predihood.selection import retrieve_lists
+from predihood.utility_functions import check_dataset_size, get_most_frequent
+from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
+from sklearn.exceptions import ConvergenceWarning
+from sklearn.linear_model import LogisticRegression, SGDClassifier
+from sklearn.neighbors import KNeighborsClassifier, NearestCentroid
+from sklearn.svm import SVC
+from sklearn.utils._testing import ignore_warnings
+
+log = logging.getLogger(__name__)
+
+
+# define classifiers used in the experimental validation
+CLASSIFIERS = [
+ LogisticRegression(penalty="l2"),
+ KNeighborsClassifier(n_neighbors=30),
+ RandomForestClassifier(n_estimators=300),
+ SVC(kernel="rbf"),
+ SGDClassifier(),
+ NearestCentroid(),
+ AdaBoostClassifier()
+]
+
+
+@ignore_warnings(category=ConvergenceWarning)
+def compute_all_accuracies(data, clf, train_size, test_size, remove_outliers=False, remove_rural=False):
+ """
+ Compute accuracies for each EV and each list with the given classifier.
+
+ Args:
+ data: a Data object that contains assessed IRIS and their attributes
+ clf: an object which is a classifier (with `fit` and `predict` methods)
+ train_size: an integer corresponding to the size of the training sample
+ test_size: an integer corresponding to the size of the test sample
+ remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
+ remove_rural: True to remove IRIS that are in the countryside to avoid bias while predicting, False else
+
+ Returns:
+ a dictionary of results for each EV and each list of selected indicators
+ """
+ log.info("... Computing accuracies ...")
+ train_size, test_size = check_dataset_size(train_size, test_size)
+
+ data_not_filtered = Data(normalization="density", filtering=False)
+ data_not_filtered.init_all_in_one()
+
+ lists = retrieve_lists()
+ results = {}
+ for j, env in enumerate(ENVIRONMENT_VARIABLES):
+ results[env] = OrderedDict()
+ log.debug("--- %s ---", env)
+
+ dataset = Dataset(data_not_filtered, env, selected_indicators=data_not_filtered.indicators, train_size=train_size, test_size=test_size, outliers=remove_outliers, _type='supervised')
+ dataset.init_all_in_one()
+ if remove_rural: dataset.remove_rural_iris()
+
+ mean_classifier = 0.0
+ algo = MethodPrediction(name="", dataset=dataset, classifier=clf)
+ algo.fit()
+ algo.compute_performance()
+ results[env]["accuracy_none"] = algo.accuracy
+ results[env]["accuracies"] = OrderedDict()
+ log.info("accuracy for %s without filtering: %f", env, algo.accuracy)
+
+ for top_k, lst in lists.items():
+ dataset = Dataset(data, env, selected_indicators=lst[env], train_size=train_size, test_size=test_size, outliers=remove_outliers, _type='supervised')
+ dataset.init_all_in_one()
+ if remove_rural: dataset.remove_rural_iris()
+ algo2 = MethodPrediction(name='', dataset=dataset, classifier=clf)
+ algo2.fit()
+ algo2.compute_performance()
+ mean_classifier += algo2.accuracy
+ results[env]["accuracies"][str(top_k)] = algo2.accuracy
+ log.info("accuracy for %s with %s: %f", env, top_k, algo2.accuracy)
+ print("means:", results[env])
+ mean_classifier /= len(CLASSIFIERS)
+ results[env]["mean"] = mean_classifier
+ log.info("mean for classifier: %f", mean_classifier)
+ results = OrderedDict(results)
+ log.info(results)
+ return results
+
+
+def predict_one_iris(iris_code, data, clf, train_size, test_size, remove_outliers=False):
+ """
+ Predict the 6 EV for the given IRIS, the given data and the given classifier.
+
+ Args:
+ iris_code: a string that contains the code of the IRIS (9 digits)
+ data: a Data object on which classifier will learn
+ clf: an object (classifier) to perform the prediction
+ train_size: an integer corresponding to the size of the train sample
+ test_size: an integer corresponding to the size of the test sample
+ remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
+
+ Returns:
+ A dictionary containing predictions for each EV.
+ """
+ train_size, test_size = check_dataset_size(train_size, test_size)
+ lists = retrieve_lists()
+
+ predictions = {}
+ for j, env in enumerate(ENVIRONMENT_VARIABLES):
+ log.info("predicting for %s", env)
+ predictions_lst = []
+ for top_k, lst in lists.items():
+ dataset = Dataset(data, env, selected_indicators=lst[env], train_size=train_size, test_size=test_size, outliers=remove_outliers, _type="supervised")
+ dataset.init_all_in_one()
+ algorithm = MethodPrediction(name='', dataset=dataset, classifier=clf)
+ algorithm.fit()
+ algorithm.predict(iris_code)
+ predictions_lst.append(algorithm.prediction)
+ predictions[env] = get_most_frequent(predictions_lst) # get the most frequent value and the number of occurrences
+ print(predictions) # TODO: give an example of the dictionary
+ return predictions
+
+
+if __name__ == '__main__':
+
+ # Create data
+ data = Data(normalization="population", filtering=True)
+ data.init_all_in_one()
+
+ # Compute accuracies for each EV and each top-k
+ compute_all_accuracies(data, RandomForestClassifier(), TRAIN_SIZE, TEST_SIZE)
+
+ # Predict EV of the "Part-Dieu" IRIS, which is he CBD of Lyon (Central Business District)
+ predict_one_iris("693830301", data, RandomForestClassifier(), TRAIN_SIZE, TEST_SIZE)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="predict.compute_all_accuracies"><code class="name flex">
+<span>def <span class="ident">compute_all_accuracies</span></span>(<span>data, clf, train_size, test_size, remove_outliers=False, remove_rural=False)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Compute accuracies for each EV and each list with the given classifier.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>data</code></strong></dt>
+<dd>a Data object that contains assessed IRIS and their attributes</dd>
+<dt><strong><code>clf</code></strong></dt>
+<dd>an object which is a classifier (with <code>fit</code> and <code><a title="predict" href="#predict">predict</a></code> methods)</dd>
+<dt><strong><code>train_size</code></strong></dt>
+<dd>an integer corresponding to the size of the training sample</dd>
+<dt><strong><code>test_size</code></strong></dt>
+<dd>an integer corresponding to the size of the test sample</dd>
+<dt><strong><code>remove_outliers</code></strong></dt>
+<dd>True to remove from the dataset IRIS that are detected as outliers, False else</dd>
+<dt><strong><code>remove_rural</code></strong></dt>
+<dd>True to remove IRIS that are in the countryside to avoid bias while predicting, False else</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a dictionary</code> of <code>results for each EV and each list</code> of <code>selected indicators</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">@ignore_warnings(category=ConvergenceWarning)
+def compute_all_accuracies(data, clf, train_size, test_size, remove_outliers=False, remove_rural=False):
+ """
+ Compute accuracies for each EV and each list with the given classifier.
+
+ Args:
+ data: a Data object that contains assessed IRIS and their attributes
+ clf: an object which is a classifier (with `fit` and `predict` methods)
+ train_size: an integer corresponding to the size of the training sample
+ test_size: an integer corresponding to the size of the test sample
+ remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
+ remove_rural: True to remove IRIS that are in the countryside to avoid bias while predicting, False else
+
+ Returns:
+ a dictionary of results for each EV and each list of selected indicators
+ """
+ log.info("... Computing accuracies ...")
+ train_size, test_size = check_dataset_size(train_size, test_size)
+
+ data_not_filtered = Data(normalization="density", filtering=False)
+ data_not_filtered.init_all_in_one()
+
+ lists = retrieve_lists()
+ results = {}
+ for j, env in enumerate(ENVIRONMENT_VARIABLES):
+ results[env] = OrderedDict()
+ log.debug("--- %s ---", env)
+
+ dataset = Dataset(data_not_filtered, env, selected_indicators=data_not_filtered.indicators, train_size=train_size, test_size=test_size, outliers=remove_outliers, _type='supervised')
+ dataset.init_all_in_one()
+ if remove_rural: dataset.remove_rural_iris()
+
+ mean_classifier = 0.0
+ algo = MethodPrediction(name="", dataset=dataset, classifier=clf)
+ algo.fit()
+ algo.compute_performance()
+ results[env]["accuracy_none"] = algo.accuracy
+ results[env]["accuracies"] = OrderedDict()
+ log.info("accuracy for %s without filtering: %f", env, algo.accuracy)
+
+ for top_k, lst in lists.items():
+ dataset = Dataset(data, env, selected_indicators=lst[env], train_size=train_size, test_size=test_size, outliers=remove_outliers, _type='supervised')
+ dataset.init_all_in_one()
+ if remove_rural: dataset.remove_rural_iris()
+ algo2 = MethodPrediction(name='', dataset=dataset, classifier=clf)
+ algo2.fit()
+ algo2.compute_performance()
+ mean_classifier += algo2.accuracy
+ results[env]["accuracies"][str(top_k)] = algo2.accuracy
+ log.info("accuracy for %s with %s: %f", env, top_k, algo2.accuracy)
+ print("means:", results[env])
+ mean_classifier /= len(CLASSIFIERS)
+ results[env]["mean"] = mean_classifier
+ log.info("mean for classifier: %f", mean_classifier)
+ results = OrderedDict(results)
+ log.info(results)
+ return results</code></pre>
+</details>
+</dd>
+<dt id="predict.predict_one_iris"><code class="name flex">
+<span>def <span class="ident">predict_one_iris</span></span>(<span>iris_code, data, clf, train_size, test_size, remove_outliers=False)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Predict the 6 EV for the given IRIS, the given data and the given classifier.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>iris_code</code></strong></dt>
+<dd>a string that contains the code of the IRIS (9 digits)</dd>
+<dt><strong><code>data</code></strong></dt>
+<dd>a Data object on which classifier will learn</dd>
+<dt><strong><code>clf</code></strong></dt>
+<dd>an object (classifier) to perform the prediction</dd>
+<dt><strong><code>train_size</code></strong></dt>
+<dd>an integer corresponding to the size of the train sample</dd>
+<dt><strong><code>test_size</code></strong></dt>
+<dd>an integer corresponding to the size of the test sample</dd>
+<dt><strong><code>remove_outliers</code></strong></dt>
+<dd>True to remove from the dataset IRIS that are detected as outliers, False else</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>A dictionary containing predictions for each EV.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def predict_one_iris(iris_code, data, clf, train_size, test_size, remove_outliers=False):
+ """
+ Predict the 6 EV for the given IRIS, the given data and the given classifier.
+
+ Args:
+ iris_code: a string that contains the code of the IRIS (9 digits)
+ data: a Data object on which classifier will learn
+ clf: an object (classifier) to perform the prediction
+ train_size: an integer corresponding to the size of the train sample
+ test_size: an integer corresponding to the size of the test sample
+ remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
+
+ Returns:
+ A dictionary containing predictions for each EV.
+ """
+ train_size, test_size = check_dataset_size(train_size, test_size)
+ lists = retrieve_lists()
+
+ predictions = {}
+ for j, env in enumerate(ENVIRONMENT_VARIABLES):
+ log.info("predicting for %s", env)
+ predictions_lst = []
+ for top_k, lst in lists.items():
+ dataset = Dataset(data, env, selected_indicators=lst[env], train_size=train_size, test_size=test_size, outliers=remove_outliers, _type="supervised")
+ dataset.init_all_in_one()
+ algorithm = MethodPrediction(name='', dataset=dataset, classifier=clf)
+ algorithm.fit()
+ algorithm.predict(iris_code)
+ predictions_lst.append(algorithm.prediction)
+ predictions[env] = get_most_frequent(predictions_lst) # get the most frequent value and the number of occurrences
+ print(predictions) # TODO: give an example of the dictionary
+ return predictions</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="predict.compute_all_accuracies" href="#predict.compute_all_accuracies">compute_all_accuracies</a></code></li>
+<li><code><a title="predict.predict_one_iris" href="#predict.predict_one_iris">predict_one_iris</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/selection.html b/doc/selection.html
new file mode 100644
index 00000000..58c8aa73
--- /dev/null
+++ b/doc/selection.html
@@ -0,0 +1,406 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>selection API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>selection</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import json
+import logging
+import numpy as np
+import os
+import pandas as pd
+
+from predihood.classes.Data import Data
+from predihood.classes.Dataset import Dataset
+from predihood.classes.MethodSelection import MethodSelection
+from predihood.config import TOPS_K, ENVIRONMENT_VARIABLES, FOLDER_SELECTED_INDICATORS, FILE_HIERARCHY
+from predihood.utility_functions import apply_hierarchy
+from sklearn.ensemble import RandomForestClassifier, ExtraTreesClassifier
+
+log = logging.getLogger(__name__)
+
+
+def generate_all_data():
+ """
+ Generate all datasets, i.e. one with density normalization, one for population normalization and one without normalization.
+ The three are filtered. Generated datasets are located in generated_files/datasets.
+ """
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ data = Data(normalization="population", filtering=True)
+ data.init_all_in_one()
+ data = Data(normalization=None, filtering=True)
+ data.init_all_in_one()
+
+
+def retrieve_lists(top_k=None):
+ """
+ Get all lists generated for each top-k or for the given top-k if provided.
+
+ Args:
+ top_k: a integer corresponding to the top-k of the list to retrieve
+
+ Returns:
+ a dictionary containing the lists of indicators for each top-k
+ """
+ lists = {}
+
+ if top_k:
+ # a top-k is specified, so the list of indicators of size top-k is retrieved
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
+ else:
+ # no top-k is provided, so all lists form 10 to 100 indicators are retrieved
+ for top_k in TOPS_K:
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
+ return lists
+
+
+def retrieve_one_list(top_k):
+ """
+ Retrieve the list of selected INSEE indicators corresponding to the given top-k.
+
+ Args:
+ top_k: an integer corresponding to the size of the list to retrieve
+
+ Returns:
+ a list containing indicators of the list of size top-k
+ """
+ if not os.path.exists(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv")): generate_lists()
+ indicators_csv = pd.read_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv"), header=None)
+ lst = json.loads(indicators_csv.drop(indicators_csv.columns[0], axis=1).to_json(orient="index"))
+ list_temp = {}
+ for key, value in lst.items():
+ list_temp[ENVIRONMENT_VARIABLES[int(key)]] = [value2 for key2, value2 in value.items() if value2 is not None]
+ return list_temp
+
+
+def generate_lists():
+ """
+ Generates lists of INSEE indicators that are relevant for prediction.
+ This selection process is based on:
+ - removing fully correlated indicators
+ - select a limited number among the most relevant indicators (using Random Forest and Extra Tree classifiers)
+ - taking into account the diversity of categories of INSEE indicators based on a hierarchy of these indicators
+ """
+ # 1. Create data
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+
+ # 2. Run heat map and get fully correlated indicators by using a correlation matrtix
+ dataset = Dataset(data, "building_type", "unsupervised")
+ dataset.init_all_in_one()
+ heat_map = MethodSelection(name="heat map", dataset=dataset, parameters={"method": "spearman"})
+ heat_map.compute_selection()
+ fully_correlated_indicators = heat_map.best_indicators
+ log.info("fully correlated indicators: %d %s", len(fully_correlated_indicators), ", ".join(fully_correlated_indicators))
+
+ # 3. Select a limited number (top-k) of indicators by using Random Forest and Extra Tree classifiers.
+ # Then take into account the diversity of indicators by using a hierarchy of the INSEE indicators
+ hierarchy = pd.read_csv(FILE_HIERARCHY, sep="\t")
+
+ # 4. For each size of list (i.e. for each top-k), select a limited number of indicators with Extra Tree and Random Forest classifiers
+ # Then merge the two results to obtain a single list of relevant indicators.
+ for top_k in TOPS_K:
+ log.info("constructing list of %d indicators", top_k)
+ all_lists = [] # to keep lists of indicators for each EV (for the current top-k)
+
+ for env in ENVIRONMENT_VARIABLES:
+ dataset = Dataset(data, env, indicators_to_remove=fully_correlated_indicators, _type="supervised")
+ dataset.init_all_in_one()
+
+ # a. get best indicators according to Extra Tree classifier
+ fi_et = MethodSelection(name="feature importance ET", dataset=dataset, classifier=ExtraTreesClassifier(), parameters={"top_k": top_k})
+ fi_et.fit()
+ fi_et.compute_selection()
+ best_indicators_ET = fi_et.best_indicators
+
+ # b. get best indicators according to Random Forest classifier
+ fi_rf = MethodSelection(name="feature importance RF", dataset=dataset, classifier=RandomForestClassifier(), parameters={"top_k": top_k})
+ fi_rf.fit()
+ fi_rf.compute_selection()
+ best_indicators_RF = fi_rf.best_indicators
+
+ # c. merge indicators that have been selected by ET and RF classifiers
+ # in this step, if an indicator have been selected by the two classifiers, its score is the addition of its score for RF and the one for ET.
+ best_indicators_ET.extend(best_indicators_RF)
+ all_selected_indicators = best_indicators_ET # all selected indicators, i.e. union between RF and ET
+ keys = set([element[0] for element in all_selected_indicators]) # store indicators' names selected
+ merged_indicators_temp = {key: 0 for key in keys}
+
+ for i in range(len(all_selected_indicators)):
+ indicator = all_selected_indicators[i]
+ merged_indicators_temp[indicator[0]] += indicator[1] # adding score
+
+ # transform it into a list of sub-lists, e.g. [[indicator1, score1], ..., indicatorN, scoreN]]
+ merged_indicators = [[key, merged_indicators_temp[key]] for key in merged_indicators_temp]
+
+ # d. apply hierarchy on selected indicators to taking into account the diversity of categories of indicators
+ indicators_hierarchy = apply_hierarchy(merged_indicators, hierarchy)
+
+ # get the names of each selected indicator
+ selected_indicators_names = [indicator[0] for indicator in indicators_hierarchy]
+
+ # e. add uncorrelated indicators of heat map to the lists
+ all_lists.append(selected_indicators_names)
+
+ # C. Transform lists of selected indicators for the current top-k and save it as a CSV file
+ # indexes = {i: ENVIRONMENT_VARIABLES[i] for i in range(len(ENVIRONMENT_VARIABLES))}
+ selected_indicators = pd.DataFrame(np.array(all_lists).tolist())
+ selected_indicators.to_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list"+str(top_k)+".csv"), header=False)
+
+
+if __name__ == "__main__":
+ generate_all_data()
+ generate_lists()</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="selection.generate_all_data"><code class="name flex">
+<span>def <span class="ident">generate_all_data</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Generate all datasets, i.e. one with density normalization, one for population normalization and one without normalization.
+The three are filtered. Generated datasets are located in generated_files/datasets.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def generate_all_data():
+ """
+ Generate all datasets, i.e. one with density normalization, one for population normalization and one without normalization.
+ The three are filtered. Generated datasets are located in generated_files/datasets.
+ """
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+ data = Data(normalization="population", filtering=True)
+ data.init_all_in_one()
+ data = Data(normalization=None, filtering=True)
+ data.init_all_in_one()</code></pre>
+</details>
+</dd>
+<dt id="selection.generate_lists"><code class="name flex">
+<span>def <span class="ident">generate_lists</span></span>(<span>)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Generates lists of INSEE indicators that are relevant for prediction.
+This selection process is based on:
+- removing fully correlated indicators
+- select a limited number among the most relevant indicators (using Random Forest and Extra Tree classifiers)
+- taking into account the diversity of categories of INSEE indicators based on a hierarchy of these indicators</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def generate_lists():
+ """
+ Generates lists of INSEE indicators that are relevant for prediction.
+ This selection process is based on:
+ - removing fully correlated indicators
+ - select a limited number among the most relevant indicators (using Random Forest and Extra Tree classifiers)
+ - taking into account the diversity of categories of INSEE indicators based on a hierarchy of these indicators
+ """
+ # 1. Create data
+ data = Data(normalization="density", filtering=True)
+ data.init_all_in_one()
+
+ # 2. Run heat map and get fully correlated indicators by using a correlation matrtix
+ dataset = Dataset(data, "building_type", "unsupervised")
+ dataset.init_all_in_one()
+ heat_map = MethodSelection(name="heat map", dataset=dataset, parameters={"method": "spearman"})
+ heat_map.compute_selection()
+ fully_correlated_indicators = heat_map.best_indicators
+ log.info("fully correlated indicators: %d %s", len(fully_correlated_indicators), ", ".join(fully_correlated_indicators))
+
+ # 3. Select a limited number (top-k) of indicators by using Random Forest and Extra Tree classifiers.
+ # Then take into account the diversity of indicators by using a hierarchy of the INSEE indicators
+ hierarchy = pd.read_csv(FILE_HIERARCHY, sep="\t")
+
+ # 4. For each size of list (i.e. for each top-k), select a limited number of indicators with Extra Tree and Random Forest classifiers
+ # Then merge the two results to obtain a single list of relevant indicators.
+ for top_k in TOPS_K:
+ log.info("constructing list of %d indicators", top_k)
+ all_lists = [] # to keep lists of indicators for each EV (for the current top-k)
+
+ for env in ENVIRONMENT_VARIABLES:
+ dataset = Dataset(data, env, indicators_to_remove=fully_correlated_indicators, _type="supervised")
+ dataset.init_all_in_one()
+
+ # a. get best indicators according to Extra Tree classifier
+ fi_et = MethodSelection(name="feature importance ET", dataset=dataset, classifier=ExtraTreesClassifier(), parameters={"top_k": top_k})
+ fi_et.fit()
+ fi_et.compute_selection()
+ best_indicators_ET = fi_et.best_indicators
+
+ # b. get best indicators according to Random Forest classifier
+ fi_rf = MethodSelection(name="feature importance RF", dataset=dataset, classifier=RandomForestClassifier(), parameters={"top_k": top_k})
+ fi_rf.fit()
+ fi_rf.compute_selection()
+ best_indicators_RF = fi_rf.best_indicators
+
+ # c. merge indicators that have been selected by ET and RF classifiers
+ # in this step, if an indicator have been selected by the two classifiers, its score is the addition of its score for RF and the one for ET.
+ best_indicators_ET.extend(best_indicators_RF)
+ all_selected_indicators = best_indicators_ET # all selected indicators, i.e. union between RF and ET
+ keys = set([element[0] for element in all_selected_indicators]) # store indicators' names selected
+ merged_indicators_temp = {key: 0 for key in keys}
+
+ for i in range(len(all_selected_indicators)):
+ indicator = all_selected_indicators[i]
+ merged_indicators_temp[indicator[0]] += indicator[1] # adding score
+
+ # transform it into a list of sub-lists, e.g. [[indicator1, score1], ..., indicatorN, scoreN]]
+ merged_indicators = [[key, merged_indicators_temp[key]] for key in merged_indicators_temp]
+
+ # d. apply hierarchy on selected indicators to taking into account the diversity of categories of indicators
+ indicators_hierarchy = apply_hierarchy(merged_indicators, hierarchy)
+
+ # get the names of each selected indicator
+ selected_indicators_names = [indicator[0] for indicator in indicators_hierarchy]
+
+ # e. add uncorrelated indicators of heat map to the lists
+ all_lists.append(selected_indicators_names)
+
+ # C. Transform lists of selected indicators for the current top-k and save it as a CSV file
+ # indexes = {i: ENVIRONMENT_VARIABLES[i] for i in range(len(ENVIRONMENT_VARIABLES))}
+ selected_indicators = pd.DataFrame(np.array(all_lists).tolist())
+ selected_indicators.to_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list"+str(top_k)+".csv"), header=False)</code></pre>
+</details>
+</dd>
+<dt id="selection.retrieve_lists"><code class="name flex">
+<span>def <span class="ident">retrieve_lists</span></span>(<span>top_k=None)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get all lists generated for each top-k or for the given top-k if provided.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>top_k</code></strong></dt>
+<dd>a integer corresponding to the top-k of the list to retrieve</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a dictionary containing the lists</code> of <code>indicators for each top-k</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def retrieve_lists(top_k=None):
+ """
+ Get all lists generated for each top-k or for the given top-k if provided.
+
+ Args:
+ top_k: a integer corresponding to the top-k of the list to retrieve
+
+ Returns:
+ a dictionary containing the lists of indicators for each top-k
+ """
+ lists = {}
+
+ if top_k:
+ # a top-k is specified, so the list of indicators of size top-k is retrieved
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
+ else:
+ # no top-k is provided, so all lists form 10 to 100 indicators are retrieved
+ for top_k in TOPS_K:
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
+ return lists</code></pre>
+</details>
+</dd>
+<dt id="selection.retrieve_one_list"><code class="name flex">
+<span>def <span class="ident">retrieve_one_list</span></span>(<span>top_k)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Retrieve the list of selected INSEE indicators corresponding to the given top-k.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>top_k</code></strong></dt>
+<dd>an integer corresponding to the size of the list to retrieve</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a list containing indicators</code> of <code>the list</code> of <code>size top-k</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def retrieve_one_list(top_k):
+ """
+ Retrieve the list of selected INSEE indicators corresponding to the given top-k.
+
+ Args:
+ top_k: an integer corresponding to the size of the list to retrieve
+
+ Returns:
+ a list containing indicators of the list of size top-k
+ """
+ if not os.path.exists(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv")): generate_lists()
+ indicators_csv = pd.read_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv"), header=None)
+ lst = json.loads(indicators_csv.drop(indicators_csv.columns[0], axis=1).to_json(orient="index"))
+ list_temp = {}
+ for key, value in lst.items():
+ list_temp[ENVIRONMENT_VARIABLES[int(key)]] = [value2 for key2, value2 in value.items() if value2 is not None]
+ return list_temp</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="selection.generate_all_data" href="#selection.generate_all_data">generate_all_data</a></code></li>
+<li><code><a title="selection.generate_lists" href="#selection.generate_lists">generate_lists</a></code></li>
+<li><code><a title="selection.retrieve_lists" href="#selection.retrieve_lists">retrieve_lists</a></code></li>
+<li><code><a title="selection.retrieve_one_list" href="#selection.retrieve_one_list">retrieve_one_list</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/doc/utility_functions.html b/doc/utility_functions.html
new file mode 100644
index 00000000..80aa0b07
--- /dev/null
+++ b/doc/utility_functions.html
@@ -0,0 +1,1264 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
+<meta name="generator" content="pdoc 0.8.1" />
+<title>utility_functions API documentation</title>
+<meta name="description" content="" />
+<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
+<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
+<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
+<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
+<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
+<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>utility_functions</code></h1>
+</header>
+<section id="section-intro">
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">import ast
+import inspect
+import json
+import logging
+import numpy as np
+import pandas as pd
+import re
+import requests
+import stringdist
+
+from area import area
+from predihood import model
+from predihood.config import AVAILABLE_CLASSIFIERS, TRAIN_SIZE, TEST_SIZE, ENVIRONMENT_VARIABLES, FOLDER_DATASETS, FOLDER_DISTRIBUTION, TRANSLATION, OLD_PREFIX, NEW_PREFIX, FILE_MANUAL_ASSESSMENT
+
+log = logging.getLogger(__name__)
+
+
+# 1. IRIS function
+def address_to_code(address):
+ """
+ Get IRIS code from address.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the code of the searched address
+ """
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
+ json_response = response.json()
+ return str(json_response["complete_code"]) if "complete_code" in json_response else None
+
+
+def address_to_city(address):
+ """
+ Get city from address.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the city of the searched address
+ """
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
+ json_response = response.json()
+ return str(json_response["name"]) if "name" in json_response else None
+
+
+def append_indicator(raw_indicator, iris, lst, append_col, indicators):
+ """
+ Append the value of an INSEE indicator to a list corresponding to the data of the assessed IRIS (departure or arrival).
+
+ Args:
+ raw_indicator: a string containing the name of the indicator, e.g. P14_POP
+ iris: the iris object that contains the indicator
+ lst: the list where to append the indicator's value
+ append_col: True to append the indicator to the cols variable
+ indicators: the list of the columns of the dataset, i.e. the indicators
+
+ Returns:
+ the list containing indicator' values and the other one containing indicators names.
+ """
+ if raw_indicator in iris["properties"]["raw_indicators"]:
+ val_indicator = iris["properties"]["raw_indicators"][raw_indicator]
+ # append NaN if indicator is null, else append the value
+ lst.append(np.nan) if (val_indicator is None or not val_indicator) else lst.append(val_indicator)
+ else:
+ lst.append(np.nan)
+ if append_col: indicators.append(raw_indicator)
+ return lst, indicators
+
+
+def append_target(row, target, lst):
+ """
+ Append the target to the current list (departure or arrival).
+
+ Args:
+ row: a list corresponding to the current row of dataset
+ target: a string containing the target to append, i.e. the value of the EV
+ lst: the list where the target is append
+
+ Returns:
+ the list with the appended target
+ """
+ if target == OLD_PREFIX + "geographical_position" or target == NEW_PREFIX + "geographical_position":
+ # get only geo position without the city, e.g. South-East Lyon gives South-East
+ if str(row[target]) == "nan":
+ city = np.nan
+ else:
+ city = row[target].split(" ")[0]
+ # city_name = ''.join(split_city[1:len(split_city)])
+ value = TRANSLATION[city] if city in TRANSLATION else np.nan
+ lst.append(value)
+ else:
+ value = TRANSLATION[row[target]] if row[target] in TRANSLATION else np.nan
+ lst.append(value)
+
+ return lst
+
+
+def indicator_full_to_short_label(full_label):
+ """
+ Convert the full label of an indicator to its short label.
+
+ Args:
+ full_label: a string containing the full label of the indicator, e.g. Population 0-2 y.o. in 2014
+
+ Returns:
+ the short label of the given indicator, e.g. P14_POP0002
+ """
+ indicators = model.get_indicators_dict()
+ key_list = list(indicators.keys())
+ val_list = list(indicators.values())
+ if full_label not in indicators.values():
+ return key_list[val_list.index(full_label + " ")]
+ else:
+ return key_list[val_list.index(full_label)]
+
+
+def indicator_short_to_full_label(short_label):
+ """
+ Convert the short label of an indicator to its full label.
+
+ Args:
+ short_label: a string containing the short label of the indicator, e.g. P14_POP0002
+
+ Returns:
+ the full label of the indicator, e.g. Population 0-2 y.o. in 2014
+ """
+ indicators = model.get_indicators_dict()
+ return indicators[short_label]
+
+
+def apply_hierarchy(selected_indicators, hierarchy):
+ """
+ Apply hierarchy on the given indicators in order to go up children indicators in their parents.
+
+ Args:
+ selected_indicators: a list of indicators selected by the feature importance process.
+ hierarchy: the hierarchy of the indicators, i.e. for each indicator, there are its level in the hierarchy and its first ancestor.
+
+ Returns:
+ the new selected indicators list, with some children which have go up in their parents.
+ """
+ list_indicators_FI = [selected_indicators[j][0] for j in range(len(selected_indicators))]
+ indexes_to_remove = []
+ for i in range(len(selected_indicators)):
+ index_row = hierarchy.index[hierarchy["INDICATOR"] == selected_indicators[i][0]].tolist()
+ if len(index_row) == 0:
+ continue # pass this indicator because it does not exist in the hierarchy
+ else:
+ index_row = index_row[0]
+ level = hierarchy.iloc[index_row, 2]
+ ancestor = hierarchy.iloc[index_row, 3]
+ if level > 1:
+ if ancestor in list_indicators_FI:
+ index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
+ while index_ancestor in indexes_to_remove:
+ ancestor2 = hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 3] # name of ancestor
+ ancestor = ancestor2
+ index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
+ if hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 2] == 1:
+ break
+ if index_ancestor not in indexes_to_remove:
+ selected_indicators[index_ancestor][1] += selected_indicators[i][1]
+ indexes_to_remove.append(i)
+ for index in sorted(indexes_to_remove,reverse=True): # remove in reverse order to do not throw off subsequent indexes
+ del selected_indicators[index]
+ return selected_indicators
+
+
+def get_classifier(name):
+ """
+ Get an instance of a classifier with its name.
+
+ Args:
+ name: a string containing the name of the desired classifier
+
+ Returns:
+ an instance of a classifier
+ """
+ classifier = AVAILABLE_CLASSIFIERS[name]
+ return classifier()
+
+
+def set_classifier(classifier, parameters):
+ """
+ Tune the classifier with the given parameters.
+
+ Args:
+ classifier: an instance of the classifier to tune
+ parameters: a dictionary containing the tuning parameters
+
+ Returns:
+ an instance of the tuned classifier
+ """
+ keys_clf = list(classifier.get_params().keys())
+ # remove None parameters and parameters that don't exist in sklearn (e.g. train and test size)
+ parameters = {key: value for key, value in parameters.items() if value != "" and key in keys_clf}
+ classifier.set_params(**parameters)
+ return classifier
+
+
+def signature(chosen_algorithm):
+ """
+ Get the signature of an algorithm, i.e. its parameters, the default values and the type of each parameter. The documentation of the algorithm must be in NumPy style.
+
+ Args:
+ chosen_algorithm: the name of the algorithm in str, e.g. RandomForestClassifier
+
+ Returns:
+ the signature of the given algorithm, i.e. a dictionary containing for each parameter:
+ - a list of the accepted types
+ - the default value
+ - a description of the parameter (e.g. "The train_size parameter aims at tuning the size of the sample during the learning step.")
+ """
+ # special case for no selection
+ if chosen_algorithm == "Algorithm": return json.dumps({})
+ try:
+ # model = eval(_chosen_algorithm) # never use eval on untrusted strings
+ model = get_classifier(chosen_algorithm)
+ doc = model.__doc__ # TODO: specify case when there is no doc (user-implemented algorithm)
+ param_section = "Parameters"
+ dashes = "-" * len(param_section) # -------
+ number_spaces = doc.find(dashes) - (doc.find(param_section) + len(param_section))
+ attribute_section = "Attributes\n"
+ # sub_doc is the param section of the docs (i.e. without attributes and some text)
+ sub_doc = doc[doc.find(param_section) + len(param_section) + number_spaces + len(dashes) + len("\n"):doc.find(attribute_section)]
+ except:
+ raise Exception("This algorithm does not exist for the moment...")
+ params = inspect.getfullargspec(model.__init__).args[1:] # get parameter' names -- [1:] to remove self parameter
+ defaults = inspect.getfullargspec(model.__init__).defaults # get default values
+ assert len(params) == len(defaults)
+ parameters = {}
+ for i in range(len(params)):
+ param_name = str(params[i]) + " : "
+ index_param = sub_doc.find(param_name)
+ index_next_newline = sub_doc[index_param:].find("\n") # find returns the first occurrence
+ parameter_string = sub_doc[index_param:index_param + index_next_newline]
+ doc_param = sub_doc[index_param + index_next_newline:]
+ index_end_sentence = re.search("(\.\s)", doc_param).start() # search for the first sentence
+ first_sentence = doc_param[:index_end_sentence + 1]
+ # format first sentence to have a prettier display.
+ first_sentence = first_sentence.replace("\n", " ")
+ while " " in first_sentence:
+ first_sentence = first_sentence.replace(" ", " ")
+ types_and_default = parameter_string[len(param_name):]
+ if "{" in types_and_default and "}" in types_and_default: # for cases like {"auto", "kd_tree", "brute"}, optional
+ types_and_default = types_and_default.replace("{", '')
+ types_and_default = types_and_default.replace("}", '')
+ if " or " in types_and_default: types_and_default = types_and_default.replace(" or ", ", ")
+ types_defaults_split = types_and_default.split(", ")
+ types = []
+ default = -1
+ variants = ["optional (default=", "optional (default = ", "optional", "(default=", "(default = ", "default ",
+ "default: ", "default="] # DO NOT CHANGE THE ORDER OF ITEMS
+ for item in types_defaults_split:
+ if not any(value in item for value in variants):
+ if item.startswith("length"):
+ pass # exceptions
+ else:
+ types.append(item) # item is a type
+ else:
+ for value in variants:
+ if value in item:
+ if value.startswith("optional ("):
+ default = item.split(value)[1][:-1]
+ elif value.startswith("(default"):
+ default = item.split(value)[1][:-1]
+ elif value.startswith("default"):
+ default = item.split(value)[1]
+ elif value == "optional":
+ default = "None"
+ break # do not iterate over other values
+ if default != -1 and default != "None":
+ type_of_default = str(type(ast.literal_eval(str(default))).__name__)
+ else:
+ type_of_default = "str"
+ types[:] = ["int" if x == "integer" else x for x in types] # replace "integer" by "int"
+ types[:] = ["bool" if x == "boolean" else x for x in types] # replace "boolean" by "bool"
+ types[:] = ["str" if x == "string" else x for x in types] # replace "str" by "string"
+ if len(types) == 0: types.append(type_of_default) # fill missing types
+ types[:] = [x for x in types if "None" not in x and "NoneType" not in x] # remove None type
+ parameters[param_name[:-3]] = {"types": types, "default": default, "description": first_sentence} # -3 to remove " : "
+ return parameters
+
+
+def check_dataset_size(train_size, test_size):
+ """
+ Check train and test size and update with defaults or divided by 100 if needed.
+
+ Args:
+ train_size: an integer or a float corresponding to the value for train size (should be between 0 and 1)
+ test_size: an integer or a float corresponding to the value for test size (should be between 0 and 1)
+
+ Returns:
+ the train and test sizes
+ """
+ if 0 < train_size < 1 and 0 < test_size < 1 and train_size + test_size == 1:
+ return train_size, test_size # default case
+ elif 0 < train_size < 1 and 0 < test_size < 1:
+ train_size = TRAIN_SIZE # 0.8
+ test_size = TEST_SIZE # 0.2
+
+ if 1 <= train_size < 100 and 1 <= test_size < 100 and train_size + test_size == 100:
+ return train_size / 100, test_size / 100 # default case
+ elif 1 <= train_size < 100 and 1 <= test_size < 100:
+ train_size = TRAIN_SIZE # 0.8
+ test_size = TEST_SIZE # 0.2
+ return train_size, test_size
+
+
+# 2. list functions
+def intersection(lst1, lst2):
+ """
+ Intersect two lists.
+
+ Args:
+ lst1: a list corresponding to the first list to be intersected
+ lst2: a list corresponding to the second list to be intersected
+
+ Returns:
+ a list corresponding to the result of the intersection of the two given lists.
+ """
+ return list(set(lst1) & set(lst2))
+
+
+def union(lst1, lst2):
+ """
+ Unify two lists without repetitions.
+
+ Args:
+ lst1: a list corresponding to the first list to append
+ lst2: a list corresponding to the second list to append
+
+ Returns:
+ a list corresponding to the union of the two lists
+ """
+ return list(set(lst1) | set(lst2))
+
+
+def similarity(value, lst):
+ """
+ Check if a value is enough similar to the data.
+
+ Args:
+ value: the value on which the similarity is computed
+ lst: the list containing other values to check similarity
+
+ Returns:
+ the index of the similar value, -1 if no value is enough similar
+ """
+ for i in range(len(lst)):
+ if value in lst[i]: return -2 # this value is already append
+ for elem in lst[i]:
+ if not isinstance(elem, float):
+ dissimilarity = stringdist.levenshtein(str(elem), str(value)) # compute Levenshtein similarity
+ if dissimilarity == 1: return i # if the given value has only one difference with the current value, store it as a similarity
+ return -1
+
+
+def get_most_frequent(lst):
+ """
+ Get the most frequent item in a list. If many elements are frequent, it returns the first one.
+
+ Args:
+ lst: the list to find the most frequent element
+
+ Returns:
+ a dictionary containing the most frequent of the given list and its count
+ """
+ most_frequent_element = max(set(lst), key=lst.count)
+ dictionary = {"most_frequent": most_frequent_element, "count_frequent": lst.count(most_frequent_element)}
+ return dictionary
+
+
+# 3. plot functions
+def auto_label(rectangles, axes):
+ """
+ Adds a text label above each bar in rectangles, displaying its value.
+
+ Args:
+ rectangles: the bars of the plot.
+ axes: the axes of the plot.
+ """
+ for rect in rectangles:
+ height = rect.get_height()
+ axes.annotate("{}".format(height), xy=(rect.get_x() + rect.get_width() / 2, height), xytext=(0, 3), textcoords="offset points", ha="center", va="bottom")
+
+
+def add_assessment_to_file(code_iris, values):
+ """
+ Add an assessed IRIS to the CSV file.
+
+ Args:
+ code_iris: a string corresponding to the code of the IRIS (9 digits)
+ values: the values of the 6 EV that represent the environment of the assessed IRIS
+
+ Returns:
+ the string "okay" if the assessed IRIS has been added to the CSV file
+ """
+ df = pd.read_csv(FILE_MANUAL_ASSESSMENT)
+ codes = df["CODE"].tolist()
+ codes_lst = [str(elem) for elem in codes]
+ if code_iris in codes_lst:
+ return "iris already assessed"
+ else:
+ iris_data = model.get_iris_from_code(code_iris)
+ iris_coords = model.get_coords_from_code(code_iris)
+ area_iris = area(iris_coords) / 1000000 if iris_coords is not None else None
+ density_iris = iris_data["properties"]["raw_indicators"]["P14_POP"] / area_iris if area_iris is not None and area_iris > 0 else None
+
+ iris = []
+ cols = ["CODE", "AREA", "DENSITY"]
+ # adding code, area and density
+ iris.append(code_iris)
+ iris.append(area_iris)
+ iris.append(density_iris)
+
+ # adding insee indicators
+ indicators = model.get_indicators_list()
+ indicators_to_remove = ["IRIS", "REG", "DEP", "UU2010", "COM", "LIBCOM", "TRIRIS", "GRD_QUART", "LIBIRIS", "TYP_IRIS", "MODIF_IRIS", "LAB_IRIS", "LIB_IRIS", "LIB_COM", "CODGEO", "LIBGEO"]
+ for indicator in indicators_to_remove:
+ if indicator in indicators:
+ indicators.remove(indicator)
+
+ for raw_indicator in indicators:
+ iris, cols = append_indicator(raw_indicator, iris_data, iris, True, cols)
+
+ iris.extend(values) # adding assessed values
+ cols.extend(ENVIRONMENT_VARIABLES)
+ df = pd.DataFrame([iris])
+ df.to_csv(FILE_MANUAL_ASSESSMENT, mode="a", index=False, header=False) # DO NOT ERASE HEADER IN THE CSV FILE
+ return "okay"</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="utility_functions.add_assessment_to_file"><code class="name flex">
+<span>def <span class="ident">add_assessment_to_file</span></span>(<span>code_iris, values)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Add an assessed IRIS to the CSV file.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>code_iris</code></strong></dt>
+<dd>a string corresponding to the code of the IRIS (9 digits)</dd>
+<dt><strong><code>values</code></strong></dt>
+<dd>the values of the 6 EV that represent the environment of the assessed IRIS</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the string "okay" if the assessed IRIS has been added to the CSV file</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def add_assessment_to_file(code_iris, values):
+ """
+ Add an assessed IRIS to the CSV file.
+
+ Args:
+ code_iris: a string corresponding to the code of the IRIS (9 digits)
+ values: the values of the 6 EV that represent the environment of the assessed IRIS
+
+ Returns:
+ the string "okay" if the assessed IRIS has been added to the CSV file
+ """
+ df = pd.read_csv(FILE_MANUAL_ASSESSMENT)
+ codes = df["CODE"].tolist()
+ codes_lst = [str(elem) for elem in codes]
+ if code_iris in codes_lst:
+ return "iris already assessed"
+ else:
+ iris_data = model.get_iris_from_code(code_iris)
+ iris_coords = model.get_coords_from_code(code_iris)
+ area_iris = area(iris_coords) / 1000000 if iris_coords is not None else None
+ density_iris = iris_data["properties"]["raw_indicators"]["P14_POP"] / area_iris if area_iris is not None and area_iris > 0 else None
+
+ iris = []
+ cols = ["CODE", "AREA", "DENSITY"]
+ # adding code, area and density
+ iris.append(code_iris)
+ iris.append(area_iris)
+ iris.append(density_iris)
+
+ # adding insee indicators
+ indicators = model.get_indicators_list()
+ indicators_to_remove = ["IRIS", "REG", "DEP", "UU2010", "COM", "LIBCOM", "TRIRIS", "GRD_QUART", "LIBIRIS", "TYP_IRIS", "MODIF_IRIS", "LAB_IRIS", "LIB_IRIS", "LIB_COM", "CODGEO", "LIBGEO"]
+ for indicator in indicators_to_remove:
+ if indicator in indicators:
+ indicators.remove(indicator)
+
+ for raw_indicator in indicators:
+ iris, cols = append_indicator(raw_indicator, iris_data, iris, True, cols)
+
+ iris.extend(values) # adding assessed values
+ cols.extend(ENVIRONMENT_VARIABLES)
+ df = pd.DataFrame([iris])
+ df.to_csv(FILE_MANUAL_ASSESSMENT, mode="a", index=False, header=False) # DO NOT ERASE HEADER IN THE CSV FILE
+ return "okay"</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.address_to_city"><code class="name flex">
+<span>def <span class="ident">address_to_city</span></span>(<span>address)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get city from address.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>address</code></strong></dt>
+<dd>a string containing the address of the iris</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a string containing the city</code> of <code>the searched address</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def address_to_city(address):
+ """
+ Get city from address.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the city of the searched address
+ """
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
+ json_response = response.json()
+ return str(json_response["name"]) if "name" in json_response else None</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.address_to_code"><code class="name flex">
+<span>def <span class="ident">address_to_code</span></span>(<span>address)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get IRIS code from address.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>address</code></strong></dt>
+<dd>a string containing the address of the iris</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a string containing the code</code> of <code>the searched address</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def address_to_code(address):
+ """
+ Get IRIS code from address.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the code of the searched address
+ """
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
+ json_response = response.json()
+ return str(json_response["complete_code"]) if "complete_code" in json_response else None</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.append_indicator"><code class="name flex">
+<span>def <span class="ident">append_indicator</span></span>(<span>raw_indicator, iris, lst, append_col, indicators)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Append the value of an INSEE indicator to a list corresponding to the data of the assessed IRIS (departure or arrival).</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>raw_indicator</code></strong></dt>
+<dd>a string containing the name of the indicator, e.g. P14_POP</dd>
+<dt><strong><code>iris</code></strong></dt>
+<dd>the iris object that contains the indicator</dd>
+<dt><strong><code>lst</code></strong></dt>
+<dd>the list where to append the indicator's value</dd>
+<dt><strong><code>append_col</code></strong></dt>
+<dd>True to append the indicator to the cols variable</dd>
+<dt><strong><code>indicators</code></strong></dt>
+<dd>the list of the columns of the dataset, i.e. the indicators</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>the list containing indicator' values and the other one containing indicators names.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def append_indicator(raw_indicator, iris, lst, append_col, indicators):
+ """
+ Append the value of an INSEE indicator to a list corresponding to the data of the assessed IRIS (departure or arrival).
+
+ Args:
+ raw_indicator: a string containing the name of the indicator, e.g. P14_POP
+ iris: the iris object that contains the indicator
+ lst: the list where to append the indicator's value
+ append_col: True to append the indicator to the cols variable
+ indicators: the list of the columns of the dataset, i.e. the indicators
+
+ Returns:
+ the list containing indicator' values and the other one containing indicators names.
+ """
+ if raw_indicator in iris["properties"]["raw_indicators"]:
+ val_indicator = iris["properties"]["raw_indicators"][raw_indicator]
+ # append NaN if indicator is null, else append the value
+ lst.append(np.nan) if (val_indicator is None or not val_indicator) else lst.append(val_indicator)
+ else:
+ lst.append(np.nan)
+ if append_col: indicators.append(raw_indicator)
+ return lst, indicators</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.append_target"><code class="name flex">
+<span>def <span class="ident">append_target</span></span>(<span>row, target, lst)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Append the target to the current list (departure or arrival).</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>row</code></strong></dt>
+<dd>a list corresponding to the current row of dataset</dd>
+<dt><strong><code>target</code></strong></dt>
+<dd>a string containing the target to append, i.e. the value of the EV</dd>
+<dt><strong><code>lst</code></strong></dt>
+<dd>the list where the target is append</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the list with the appended target</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def append_target(row, target, lst):
+ """
+ Append the target to the current list (departure or arrival).
+
+ Args:
+ row: a list corresponding to the current row of dataset
+ target: a string containing the target to append, i.e. the value of the EV
+ lst: the list where the target is append
+
+ Returns:
+ the list with the appended target
+ """
+ if target == OLD_PREFIX + "geographical_position" or target == NEW_PREFIX + "geographical_position":
+ # get only geo position without the city, e.g. South-East Lyon gives South-East
+ if str(row[target]) == "nan":
+ city = np.nan
+ else:
+ city = row[target].split(" ")[0]
+ # city_name = ''.join(split_city[1:len(split_city)])
+ value = TRANSLATION[city] if city in TRANSLATION else np.nan
+ lst.append(value)
+ else:
+ value = TRANSLATION[row[target]] if row[target] in TRANSLATION else np.nan
+ lst.append(value)
+
+ return lst</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.apply_hierarchy"><code class="name flex">
+<span>def <span class="ident">apply_hierarchy</span></span>(<span>selected_indicators, hierarchy)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Apply hierarchy on the given indicators in order to go up children indicators in their parents.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>selected_indicators</code></strong></dt>
+<dd>a list of indicators selected by the feature importance process.</dd>
+<dt><strong><code>hierarchy</code></strong></dt>
+<dd>the hierarchy of the indicators, i.e. for each indicator, there are its level in the hierarchy and its first ancestor.</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>the new selected indicators list, with some children which have go up in their parents.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def apply_hierarchy(selected_indicators, hierarchy):
+ """
+ Apply hierarchy on the given indicators in order to go up children indicators in their parents.
+
+ Args:
+ selected_indicators: a list of indicators selected by the feature importance process.
+ hierarchy: the hierarchy of the indicators, i.e. for each indicator, there are its level in the hierarchy and its first ancestor.
+
+ Returns:
+ the new selected indicators list, with some children which have go up in their parents.
+ """
+ list_indicators_FI = [selected_indicators[j][0] for j in range(len(selected_indicators))]
+ indexes_to_remove = []
+ for i in range(len(selected_indicators)):
+ index_row = hierarchy.index[hierarchy["INDICATOR"] == selected_indicators[i][0]].tolist()
+ if len(index_row) == 0:
+ continue # pass this indicator because it does not exist in the hierarchy
+ else:
+ index_row = index_row[0]
+ level = hierarchy.iloc[index_row, 2]
+ ancestor = hierarchy.iloc[index_row, 3]
+ if level > 1:
+ if ancestor in list_indicators_FI:
+ index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
+ while index_ancestor in indexes_to_remove:
+ ancestor2 = hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 3] # name of ancestor
+ ancestor = ancestor2
+ index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
+ if hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 2] == 1:
+ break
+ if index_ancestor not in indexes_to_remove:
+ selected_indicators[index_ancestor][1] += selected_indicators[i][1]
+ indexes_to_remove.append(i)
+ for index in sorted(indexes_to_remove,reverse=True): # remove in reverse order to do not throw off subsequent indexes
+ del selected_indicators[index]
+ return selected_indicators</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.auto_label"><code class="name flex">
+<span>def <span class="ident">auto_label</span></span>(<span>rectangles, axes)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Adds a text label above each bar in rectangles, displaying its value.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>rectangles</code></strong></dt>
+<dd>the bars of the plot.</dd>
+<dt><strong><code>axes</code></strong></dt>
+<dd>the axes of the plot.</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def auto_label(rectangles, axes):
+ """
+ Adds a text label above each bar in rectangles, displaying its value.
+
+ Args:
+ rectangles: the bars of the plot.
+ axes: the axes of the plot.
+ """
+ for rect in rectangles:
+ height = rect.get_height()
+ axes.annotate("{}".format(height), xy=(rect.get_x() + rect.get_width() / 2, height), xytext=(0, 3), textcoords="offset points", ha="center", va="bottom")</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.check_dataset_size"><code class="name flex">
+<span>def <span class="ident">check_dataset_size</span></span>(<span>train_size, test_size)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Check train and test size and update with defaults or divided by 100 if needed.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>train_size</code></strong></dt>
+<dd>an integer or a float corresponding to the value for train size (should be between 0 and 1)</dd>
+<dt><strong><code>test_size</code></strong></dt>
+<dd>an integer or a float corresponding to the value for test size (should be between 0 and 1)</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the train and test sizes</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def check_dataset_size(train_size, test_size):
+ """
+ Check train and test size and update with defaults or divided by 100 if needed.
+
+ Args:
+ train_size: an integer or a float corresponding to the value for train size (should be between 0 and 1)
+ test_size: an integer or a float corresponding to the value for test size (should be between 0 and 1)
+
+ Returns:
+ the train and test sizes
+ """
+ if 0 < train_size < 1 and 0 < test_size < 1 and train_size + test_size == 1:
+ return train_size, test_size # default case
+ elif 0 < train_size < 1 and 0 < test_size < 1:
+ train_size = TRAIN_SIZE # 0.8
+ test_size = TEST_SIZE # 0.2
+
+ if 1 <= train_size < 100 and 1 <= test_size < 100 and train_size + test_size == 100:
+ return train_size / 100, test_size / 100 # default case
+ elif 1 <= train_size < 100 and 1 <= test_size < 100:
+ train_size = TRAIN_SIZE # 0.8
+ test_size = TEST_SIZE # 0.2
+ return train_size, test_size</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.get_classifier"><code class="name flex">
+<span>def <span class="ident">get_classifier</span></span>(<span>name)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get an instance of a classifier with its name.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>name</code></strong></dt>
+<dd>a string containing the name of the desired classifier</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>an instance</code> of <code>a classifier</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_classifier(name):
+ """
+ Get an instance of a classifier with its name.
+
+ Args:
+ name: a string containing the name of the desired classifier
+
+ Returns:
+ an instance of a classifier
+ """
+ classifier = AVAILABLE_CLASSIFIERS[name]
+ return classifier()</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.get_most_frequent"><code class="name flex">
+<span>def <span class="ident">get_most_frequent</span></span>(<span>lst)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get the most frequent item in a list. If many elements are frequent, it returns the first one.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>lst</code></strong></dt>
+<dd>the list to find the most frequent element</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a dictionary containing the most frequent</code> of <code>the given list and its count</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def get_most_frequent(lst):
+ """
+ Get the most frequent item in a list. If many elements are frequent, it returns the first one.
+
+ Args:
+ lst: the list to find the most frequent element
+
+ Returns:
+ a dictionary containing the most frequent of the given list and its count
+ """
+ most_frequent_element = max(set(lst), key=lst.count)
+ dictionary = {"most_frequent": most_frequent_element, "count_frequent": lst.count(most_frequent_element)}
+ return dictionary</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.indicator_full_to_short_label"><code class="name flex">
+<span>def <span class="ident">indicator_full_to_short_label</span></span>(<span>full_label)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Convert the full label of an indicator to its short label.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>full_label</code></strong></dt>
+<dd>a string containing the full label of the indicator, e.g. Population 0-2 y.o. in 2014</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the short label</code> of <code>the given indicator, e.g. P14_POP0002</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def indicator_full_to_short_label(full_label):
+ """
+ Convert the full label of an indicator to its short label.
+
+ Args:
+ full_label: a string containing the full label of the indicator, e.g. Population 0-2 y.o. in 2014
+
+ Returns:
+ the short label of the given indicator, e.g. P14_POP0002
+ """
+ indicators = model.get_indicators_dict()
+ key_list = list(indicators.keys())
+ val_list = list(indicators.values())
+ if full_label not in indicators.values():
+ return key_list[val_list.index(full_label + " ")]
+ else:
+ return key_list[val_list.index(full_label)]</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.indicator_short_to_full_label"><code class="name flex">
+<span>def <span class="ident">indicator_short_to_full_label</span></span>(<span>short_label)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Convert the short label of an indicator to its full label.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>short_label</code></strong></dt>
+<dd>a string containing the short label of the indicator, e.g. P14_POP0002</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the full label</code> of <code>the indicator, e.g. Population 0-2 y.o. in 2014</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def indicator_short_to_full_label(short_label):
+ """
+ Convert the short label of an indicator to its full label.
+
+ Args:
+ short_label: a string containing the short label of the indicator, e.g. P14_POP0002
+
+ Returns:
+ the full label of the indicator, e.g. Population 0-2 y.o. in 2014
+ """
+ indicators = model.get_indicators_dict()
+ return indicators[short_label]</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.intersection"><code class="name flex">
+<span>def <span class="ident">intersection</span></span>(<span>lst1, lst2)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Intersect two lists.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>lst1</code></strong></dt>
+<dd>a list corresponding to the first list to be intersected</dd>
+<dt><strong><code>lst2</code></strong></dt>
+<dd>a list corresponding to the second list to be intersected</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>a list corresponding to the result of the intersection of the two given lists.</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def intersection(lst1, lst2):
+ """
+ Intersect two lists.
+
+ Args:
+ lst1: a list corresponding to the first list to be intersected
+ lst2: a list corresponding to the second list to be intersected
+
+ Returns:
+ a list corresponding to the result of the intersection of the two given lists.
+ """
+ return list(set(lst1) & set(lst2))</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.set_classifier"><code class="name flex">
+<span>def <span class="ident">set_classifier</span></span>(<span>classifier, parameters)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Tune the classifier with the given parameters.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>classifier</code></strong></dt>
+<dd>an instance of the classifier to tune</dd>
+<dt><strong><code>parameters</code></strong></dt>
+<dd>a dictionary containing the tuning parameters</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>an instance</code> of <code>the tuned classifier</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def set_classifier(classifier, parameters):
+ """
+ Tune the classifier with the given parameters.
+
+ Args:
+ classifier: an instance of the classifier to tune
+ parameters: a dictionary containing the tuning parameters
+
+ Returns:
+ an instance of the tuned classifier
+ """
+ keys_clf = list(classifier.get_params().keys())
+ # remove None parameters and parameters that don't exist in sklearn (e.g. train and test size)
+ parameters = {key: value for key, value in parameters.items() if value != "" and key in keys_clf}
+ classifier.set_params(**parameters)
+ return classifier</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.signature"><code class="name flex">
+<span>def <span class="ident">signature</span></span>(<span>chosen_algorithm)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Get the signature of an algorithm, i.e. its parameters, the default values and the type of each parameter. The documentation of the algorithm must be in NumPy style.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>chosen_algorithm</code></strong></dt>
+<dd>the name of the algorithm in str, e.g. RandomForestClassifier</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<p>the signature of the given algorithm, i.e. a dictionary containing for each parameter:
+- a list of the accepted types
+- the default value
+- a description of the parameter (e.g. "The train_size parameter aims at tuning the size of the sample during the learning step.")</p></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def signature(chosen_algorithm):
+ """
+ Get the signature of an algorithm, i.e. its parameters, the default values and the type of each parameter. The documentation of the algorithm must be in NumPy style.
+
+ Args:
+ chosen_algorithm: the name of the algorithm in str, e.g. RandomForestClassifier
+
+ Returns:
+ the signature of the given algorithm, i.e. a dictionary containing for each parameter:
+ - a list of the accepted types
+ - the default value
+ - a description of the parameter (e.g. "The train_size parameter aims at tuning the size of the sample during the learning step.")
+ """
+ # special case for no selection
+ if chosen_algorithm == "Algorithm": return json.dumps({})
+ try:
+ # model = eval(_chosen_algorithm) # never use eval on untrusted strings
+ model = get_classifier(chosen_algorithm)
+ doc = model.__doc__ # TODO: specify case when there is no doc (user-implemented algorithm)
+ param_section = "Parameters"
+ dashes = "-" * len(param_section) # -------
+ number_spaces = doc.find(dashes) - (doc.find(param_section) + len(param_section))
+ attribute_section = "Attributes\n"
+ # sub_doc is the param section of the docs (i.e. without attributes and some text)
+ sub_doc = doc[doc.find(param_section) + len(param_section) + number_spaces + len(dashes) + len("\n"):doc.find(attribute_section)]
+ except:
+ raise Exception("This algorithm does not exist for the moment...")
+ params = inspect.getfullargspec(model.__init__).args[1:] # get parameter' names -- [1:] to remove self parameter
+ defaults = inspect.getfullargspec(model.__init__).defaults # get default values
+ assert len(params) == len(defaults)
+ parameters = {}
+ for i in range(len(params)):
+ param_name = str(params[i]) + " : "
+ index_param = sub_doc.find(param_name)
+ index_next_newline = sub_doc[index_param:].find("\n") # find returns the first occurrence
+ parameter_string = sub_doc[index_param:index_param + index_next_newline]
+ doc_param = sub_doc[index_param + index_next_newline:]
+ index_end_sentence = re.search("(\.\s)", doc_param).start() # search for the first sentence
+ first_sentence = doc_param[:index_end_sentence + 1]
+ # format first sentence to have a prettier display.
+ first_sentence = first_sentence.replace("\n", " ")
+ while " " in first_sentence:
+ first_sentence = first_sentence.replace(" ", " ")
+ types_and_default = parameter_string[len(param_name):]
+ if "{" in types_and_default and "}" in types_and_default: # for cases like {"auto", "kd_tree", "brute"}, optional
+ types_and_default = types_and_default.replace("{", '')
+ types_and_default = types_and_default.replace("}", '')
+ if " or " in types_and_default: types_and_default = types_and_default.replace(" or ", ", ")
+ types_defaults_split = types_and_default.split(", ")
+ types = []
+ default = -1
+ variants = ["optional (default=", "optional (default = ", "optional", "(default=", "(default = ", "default ",
+ "default: ", "default="] # DO NOT CHANGE THE ORDER OF ITEMS
+ for item in types_defaults_split:
+ if not any(value in item for value in variants):
+ if item.startswith("length"):
+ pass # exceptions
+ else:
+ types.append(item) # item is a type
+ else:
+ for value in variants:
+ if value in item:
+ if value.startswith("optional ("):
+ default = item.split(value)[1][:-1]
+ elif value.startswith("(default"):
+ default = item.split(value)[1][:-1]
+ elif value.startswith("default"):
+ default = item.split(value)[1]
+ elif value == "optional":
+ default = "None"
+ break # do not iterate over other values
+ if default != -1 and default != "None":
+ type_of_default = str(type(ast.literal_eval(str(default))).__name__)
+ else:
+ type_of_default = "str"
+ types[:] = ["int" if x == "integer" else x for x in types] # replace "integer" by "int"
+ types[:] = ["bool" if x == "boolean" else x for x in types] # replace "boolean" by "bool"
+ types[:] = ["str" if x == "string" else x for x in types] # replace "str" by "string"
+ if len(types) == 0: types.append(type_of_default) # fill missing types
+ types[:] = [x for x in types if "None" not in x and "NoneType" not in x] # remove None type
+ parameters[param_name[:-3]] = {"types": types, "default": default, "description": first_sentence} # -3 to remove " : "
+ return parameters</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.similarity"><code class="name flex">
+<span>def <span class="ident">similarity</span></span>(<span>value, lst)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Check if a value is enough similar to the data.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>value</code></strong></dt>
+<dd>the value on which the similarity is computed</dd>
+<dt><strong><code>lst</code></strong></dt>
+<dd>the list containing other values to check similarity</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>the index</code> of <code>the similar value, -1 if no value is enough similar</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def similarity(value, lst):
+ """
+ Check if a value is enough similar to the data.
+
+ Args:
+ value: the value on which the similarity is computed
+ lst: the list containing other values to check similarity
+
+ Returns:
+ the index of the similar value, -1 if no value is enough similar
+ """
+ for i in range(len(lst)):
+ if value in lst[i]: return -2 # this value is already append
+ for elem in lst[i]:
+ if not isinstance(elem, float):
+ dissimilarity = stringdist.levenshtein(str(elem), str(value)) # compute Levenshtein similarity
+ if dissimilarity == 1: return i # if the given value has only one difference with the current value, store it as a similarity
+ return -1</code></pre>
+</details>
+</dd>
+<dt id="utility_functions.union"><code class="name flex">
+<span>def <span class="ident">union</span></span>(<span>lst1, lst2)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Unify two lists without repetitions.</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>lst1</code></strong></dt>
+<dd>a list corresponding to the first list to append</dd>
+<dt><strong><code>lst2</code></strong></dt>
+<dd>a list corresponding to the second list to append</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>a list corresponding to the <a title="utility_functions.union" href="#utility_functions.union">union()</a></code> of <code>the two lists</code></dt>
+<dd> </dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+</summary>
+<pre><code class="python">def union(lst1, lst2):
+ """
+ Unify two lists without repetitions.
+
+ Args:
+ lst1: a list corresponding to the first list to append
+ lst2: a list corresponding to the second list to append
+
+ Returns:
+ a list corresponding to the union of the two lists
+ """
+ return list(set(lst1) | set(lst2))</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+</section>
+</article>
+<nav id="sidebar">
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="utility_functions.add_assessment_to_file" href="#utility_functions.add_assessment_to_file">add_assessment_to_file</a></code></li>
+<li><code><a title="utility_functions.address_to_city" href="#utility_functions.address_to_city">address_to_city</a></code></li>
+<li><code><a title="utility_functions.address_to_code" href="#utility_functions.address_to_code">address_to_code</a></code></li>
+<li><code><a title="utility_functions.append_indicator" href="#utility_functions.append_indicator">append_indicator</a></code></li>
+<li><code><a title="utility_functions.append_target" href="#utility_functions.append_target">append_target</a></code></li>
+<li><code><a title="utility_functions.apply_hierarchy" href="#utility_functions.apply_hierarchy">apply_hierarchy</a></code></li>
+<li><code><a title="utility_functions.auto_label" href="#utility_functions.auto_label">auto_label</a></code></li>
+<li><code><a title="utility_functions.check_dataset_size" href="#utility_functions.check_dataset_size">check_dataset_size</a></code></li>
+<li><code><a title="utility_functions.get_classifier" href="#utility_functions.get_classifier">get_classifier</a></code></li>
+<li><code><a title="utility_functions.get_most_frequent" href="#utility_functions.get_most_frequent">get_most_frequent</a></code></li>
+<li><code><a title="utility_functions.indicator_full_to_short_label" href="#utility_functions.indicator_full_to_short_label">indicator_full_to_short_label</a></code></li>
+<li><code><a title="utility_functions.indicator_short_to_full_label" href="#utility_functions.indicator_short_to_full_label">indicator_short_to_full_label</a></code></li>
+<li><code><a title="utility_functions.intersection" href="#utility_functions.intersection">intersection</a></code></li>
+<li><code><a title="utility_functions.set_classifier" href="#utility_functions.set_classifier">set_classifier</a></code></li>
+<li><code><a title="utility_functions.signature" href="#utility_functions.signature">signature</a></code></li>
+<li><code><a title="utility_functions.similarity" href="#utility_functions.similarity">similarity</a></code></li>
+<li><code><a title="utility_functions.union" href="#utility_functions.union">union</a></code></li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</a>.</p>
+</footer>
+<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
+<script>hljs.initHighlightingOnLoad()</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/predihood/classes/Dataset.py b/predihood/classes/Dataset.py
index 0b002155..b3bbfbf8 100644
--- a/predihood/classes/Dataset.py
+++ b/predihood/classes/Dataset.py
@@ -2,6 +2,7 @@ import logging
import pandas as pd
from predihood.config import TRAIN_SIZE, TEST_SIZE, ENVIRONMENT_VARIABLES, RANDOM_STATE, FILE_ENV
+from predihood.utility_functions import check_dataset_size
from sklearn.ensemble import IsolationForest
from sklearn.model_selection import train_test_split
@@ -9,23 +10,23 @@ log = logging.getLogger(__name__)
class Dataset:
+ """
+ This class represents assessed IRIS with their indicators ans EV values. There are options, such as removing outliers or rural IRIS.
+ """
def __init__(self, data, env, _type, selected_indicators=None, indicators_to_remove=None, train_size=TRAIN_SIZE, test_size=TEST_SIZE, outliers=False):
"""
Constructor of the Dataset class. Initialize attributes.
Args:
data: an instance of Data class. Don"t forget to initialize data after created it with "init_all_in_one()" method
- env: a string representing the environment variable. Must be in ["batiment", "usage", "paysage", "morpho", "geo", "social"]
+ env: a string representing the EV, i.e. a value in ["building_type", "building_usage", "landscape", "morphological_position", "geographical_position", "social_class"]
selected_indicators: a list containing the indicators to keep in the dataset
indicators_to_remove: a list containing the indicators to remove in the dataset
- train_size: percentage of the dataset used for training. Must be between > 0 and < 1 (e.g. 0.8 represents 80% of the dataset). The sum between train and test sizes must equal to 1
- test_size: percentage of the dataset user for testing. Must be between > 0 and < 1 (e.g. 0.2 represents 20% of the dataset). The sum between train and test sizes must equal to 1
- outliers: True or False to remove outliers from dataset. Outliers are detected with IsolationForest algorithm
- """
- # if not isinstance(data, Data): raise Exception("Dataset must be an instance of Data")
- # else: self.dataset = data # an instance of Data
- # self.data_ref = data # store a copy oof the data object, needed for prediction
- self.data = data.data[:] # copy by value, not by reference
+ train_size: a integer or a float corresponding to the size of the dataset used for training
+ test_size: a integer or a float corresponding to the size of the dataset used for test
+ outliers: True or False to remove outliers from dataset (detected with IsolationForest algorithm)
+ """
+ self.data = data.data[:] # data must be a Data object
self.indicators = data.indicators[:]
self.filtering = data.filtering
self.normalization = data.normalization
@@ -38,34 +39,28 @@ class Dataset:
self.Y_train = None
self.X_test = None
self.Y_test = None
- if env in ENVIRONMENT_VARIABLES:
- self.env = env
- else:
- self.env = "batiment"
- self.train_size = train_size
- self.test_size = test_size
+ if env in ENVIRONMENT_VARIABLES: self.env = env
+ else: self.env = "building_type"
+ self.train_size, self.test_size = check_dataset_size(train_size, test_size)
self.outliers = outliers
def init_all_in_one(self):
"""
- Initialize the dataset by initializing X and Y and generating X_train, Y_train, X_test, Y_test and removing outliers if needed
+ Initialize the dataset by initializing X and Y ; generating X_train, Y_train, X_test, Y_test ; removing outliers if needed.
+ When the type is "unsupervised", split data into X and Y is not relevant (as there is no train/test sets).
"""
if self.type == "supervised":
+ if self.outliers:
+ self.remove_outliers()
self.init_X()
self.init_Y()
self.train_test()
- if self.outliers:
- self.remove_outliers()
- self.init_X()
- self.init_Y()
- self.train_test()
def init_X(self):
"""
Initialize self.X by getting indicators in dataset.
"""
assert self.data is not None
-
if self.selected_indicators is not None: # select given indicators
self.X = self.data.loc[:, self.selected_indicators]
if self.indicators_to_remove: # remove given indicators
@@ -83,9 +78,9 @@ class Dataset:
def init_Y(self):
"""
- Initialize self.Y by getting environmental variables in dataset.
+ Initialize self.Y by getting EV in dataset.
"""
- if self.data is None: raise Exception("You must have a non empty dataset") # dataset.read()
+ assert self.data is not None
self.Y = self.data[self.env].values
def train_test(self):
@@ -100,14 +95,14 @@ class Dataset:
"""
Detect and remove IRIS that are outliers from dataset.
"""
- isolation_forest = IsolationForest(random_state=RANDOM_STATE, n_estimators=100)
+ isolation_forest = IsolationForest(random_state=RANDOM_STATE, n_estimators=100) # IsolationForest is used to detect outliers
isolation_forest.fit(self.X_train)
predictions = isolation_forest.predict(self.X_test)
for i in range(len(predictions)):
if predictions[i] == -1:
code = self.data.loc[i, "CODE"]
log.debug(code, "has been removed since it is an outlier.")
- self.data = self.data.drop(i, axis=0) # delete rows which are detected as outliers
+ self.data = self.data.drop(i, axis=0) # delete IRIS that are detected as outliers
def remove_rural_iris(self):
"""
@@ -117,19 +112,17 @@ class Dataset:
def get_environment_variable(self):
"""
- Get information about environment variable
+ Get values for a given EV for each assessed IRIS and store it in a CSV file.
"""
- if self.env not in ENVIRONMENT_VARIABLES:
- return None
- else:
- data_env = pd.DataFrame(self.data[['CODE', self.env]])
- data_env.to_csv(FILE_ENV)
+ assert self.env in ENVIRONMENT_VARIABLES
+ data_env = pd.DataFrame(self.data[['CODE', self.env]])
+ data_env.to_csv(FILE_ENV)
def get_all_environmental_variable(self):
+ """
+ Get EV for each assessed IRIS and store it in a CSV file.
+ """
columns = ['CODE']
columns.extend(ENVIRONMENT_VARIABLES)
data_env = pd.DataFrame(self.data[columns])
data_env.to_csv(FILE_ENV)
-
- def __str__(self):
- return "Class Dataset: " + "train_size=" + str(self.train_size) + " test_size=" + str(self.test_size)
diff --git a/predihood/classes/Method.py b/predihood/classes/Method.py
index 33da2dfe..e227dd76 100644
--- a/predihood/classes/Method.py
+++ b/predihood/classes/Method.py
@@ -2,17 +2,27 @@ from predihood.config import RANDOM_STATE
class Method:
+ """
+ This class represents the general concept of a Method that is applied on data. There are two specific concepts of Method: MethodSelection and MethodPrediction.
+ """
def __init__(self, name, dataset=None, classifier=None):
+ """
+ Constructor of the Method class. Initialize attributes.
+ Args:
+ name: a string that represents the name of the method, e.g. "feature selection" or "correlation matrix"
+ dataset: a Dataset object on which the method will be applied
+ classifier: an object that can be used (e.g. fit) on data
+ """
self.name = name
self.dataset = dataset
self.classifier = classifier
- self.params = None # same as return
+ self.parameters = None # same as return
def fit(self):
"""
- Fit the classifier on train set.
+ Fit the classifier on dataset.
"""
- self.classifier.random_state = RANDOM_STATE
+ self.classifier.random_state = RANDOM_STATE # defined random_state to have reproducibility
if self.dataset.type == "supervised":
self.classifier.fit(X=self.dataset.X_train, y=self.dataset.Y_train)
else:
diff --git a/predihood/classes/MethodCleaning.py b/predihood/classes/MethodCleaning.py
index 3ed30cba..49a42fd4 100644
--- a/predihood/classes/MethodCleaning.py
+++ b/predihood/classes/MethodCleaning.py
@@ -8,21 +8,24 @@ import pandas as pd
from predihood.classes.Method import Method
from predihood.config import ENVIRONMENT_VARIABLES, FILE_CLEANED_DATA, FOLDER_DISTRIBUTION, OLD_PREFIX, NEW_PREFIX
-from predihood.utility_functions import sim, auto_label
+from predihood.utility_functions import similarity, auto_label
log = logging.getLogger(__name__)
warnings.simplefilter(action='ignore', category=FutureWarning)
class MethodCleaning(Method):
+ """
+ This class represents the method for cleaning data given by the French company.
+ """
def __init__(self, name, dataset):
"""
Constructor of the MethodCleansing class. Initialize attributes.
"""
Method.__init__(self, name, dataset)
- self.values_by_env = {} # dict to store values for each environment variable, e.g. [Maisons, Maison]
- self.columns_dep = [] # departure columns, e.g. Abatiment, Ausage,...
- self.columns_arr = [] # arrival columns, e.g. Nbatiment, Nusage, ...
+ self.values_by_env = {} # dictionary to store values for each EV, e.g. [House, Houses]
+ self.columns_dep = [] # departure columns, e.g. old_building_type, old_building_usage,...
+ self.columns_arr = [] # arrival columns, e.g. new_building_type, new_building_usage, ...
self.columns = {
"occupation": {
"name_dep": "old_occupation",
@@ -50,27 +53,29 @@ class MethodCleaning(Method):
# plot variables
self.before = {} # departure values for each EV
self.after = {} # arrival values for each EV
- self.labels = {} # labels for EV, e.g. maisons, immeubles, grand ensemble...
+ self.labels = {} # labels for EV, e.g. urban, green areas, forest and country-side for the landscape variable
def clean(self):
"""
- Clean data from bad naming conventions.
+ Clean data from bad naming conventions. The idea of this function is to create a dictionary with all spellings for each value of each EV, e.g. [["Houses", "House"], ["Upper middle", "upper middle", "upper midddle"], ["Green areas"], ["Countryside"]].
+ This dictionary is constructed by computing similarities between each values and store each spelling and finally let the user choose the best one.
"""
- log.info("The data needs to be cleaned. For each list, write the correct word. For each environment variable, you will get its number of corrections and its error rate.")
- # 1. getting wrong values in a dict ordered by env variable
+ #
+ log.info("The data needs to be cleaned. For each list, write the correct word. For each EV, you will get its number of corrections and its error rate.")
+ # 1. getting wrong values in a dictionary ordered by env variable
self.values_by_env = {}
for col_dep, col_arr in zip(self.columns_dep, self.columns_arr):
col_name = col_dep.name[len(OLD_PREFIX):]
self.values_by_env[col_name] = []
for val in col_dep.unique(): # get possible values for the current column
- index = sim(val, self.values_by_env[col_name])
+ index = similarity(val, self.values_by_env[col_name])
# if the value is similar to another, add it, else create an new array with it
if index >= 0:
self.values_by_env[col_name][index].append(val)
elif index == -1:
self.values_by_env[col_name].append([val])
for val in col_arr.unique():
- index = sim(val, self.values_by_env[col_name])
+ index = similarity(val, self.values_by_env[col_name])
if index >= 0:
self.values_by_env[col_name][index].append(val)
elif index == -1:
@@ -94,7 +99,7 @@ class MethodCleaning(Method):
self.dataset.loc[self.dataset[col_name_new] == label, col_name_new] = chosen_label
size = int(self.dataset.count()[OLD_PREFIX + key]) + int(self.dataset.count()[NEW_PREFIX + key])
mean_error = ((nb_replacement_dep + nb_replacement_arr) / size) * 100
- log.debug("%d IRIS have been corrected for the environment variable %s, corresponding to an error rate of %.0f %%", (nb_replacement_dep + nb_replacement_arr), key, mean_error)
+ log.debug("%d IRIS have been corrected for the EV %s, corresponding to an error rate of %.0f %%", (nb_replacement_dep + nb_replacement_arr), key, mean_error)
# 3. removing outliers from data
count = 0
@@ -109,9 +114,11 @@ class MethodCleaning(Method):
def create_before_after_labels(self, name_dep, name_arr):
"""
- Creates the arrays 'before', 'after' and 'labels' from data.
- :param name_dep: the name of the departure column, e.g. Aoccup, Abatiment, Ausage...
- :param name_arr: the name of the arrival column, e.g. Noccup, Nbatiment, Nusage...
+ Creates the lists 'before', 'after' and 'labels' from data.
+
+ Args:
+ name_dep: a string containing the name of the departure column, e.g. old_building_type, old_building_usage...
+ name_arr: a string containing the name of the arrival column, e.g. new_building_type, new_building_usage...
"""
all_repartition = {}
self.before = {}
@@ -137,7 +144,7 @@ class MethodCleaning(Method):
else:
self.after[status] = value # self.dataset[values_after].value_counts()[status]
- # 2. merge before and after data in the sale dict
+ # 2. merge before and after data in the same dictionary
for status in self.before:
all_repartition[status] = [self.before[status], 0]
for status in self.after:
@@ -146,7 +153,7 @@ class MethodCleaning(Method):
else:
all_repartition[status][1] = self.after[status]
- # 3. convert dict in 3 arrays
+ # 3. convert dictionary in 3 arrays
self.before = []
self.after = []
self.labels = []
@@ -159,8 +166,10 @@ class MethodCleaning(Method):
def create_bar_chart(self, name, title):
"""
Plot before/after charts.
- :param name: the name of the target to plot, i.e. environment variable, e.g. usage, batiment, ...
- :param title: the title of the plot.
+
+ Args:
+ name: a string containing the name of the EV to plot, e.g. building_type, building_usage, landscape, ...
+ title: a string containing the title of the plot
"""
x = np.arange(len(self.labels)) # the label locations
width = 0.35
@@ -186,9 +195,13 @@ class MethodCleaning(Method):
def to_chart(self, env, name, title):
"""
Create before/after data and plot it.
- :param env: the target to plot, i.e. the environment variable, e.g. usage, paysage...
- :param name: the name to save the file.
- :param title: the title of the plot.
+ :param env:
+ :param name:
+ :param title:
+ Args:
+ env: a string containing the EV to plot, e.g. building_type, building_usage, landscape...
+ name: a string containing the name to save the file
+ title: a string containing the title of the plot
"""
self.create_before_after_labels(self.columns[env]["name_dep"], self.columns[env]["name_arr"])
self.create_bar_chart(name, title)
diff --git a/predihood/classes/MethodPrediction.py b/predihood/classes/MethodPrediction.py
index 8516c8c4..34d96f37 100644
--- a/predihood/classes/MethodPrediction.py
+++ b/predihood/classes/MethodPrediction.py
@@ -14,6 +14,9 @@ log = logging.getLogger(__name__)
class MethodPrediction(Method):
+ """
+ This class represents a method for predicting EV and compute performance at a national level.
+ """
def __init__(self, name, dataset, classifier):
"""
Constructor of the MethodPrediction class. Initialize attributes.
@@ -27,14 +30,14 @@ class MethodPrediction(Method):
def compute_performance(self):
"""
- Compute performance metrics like accuracy, confusion matrix, ...
+ Compute performance metrics, i.e. accuracy.
"""
scores = cross_val_score(self.classifier, self.dataset.X, self.dataset.Y, cv=5)
self.accuracy = scores.mean() * 100
def predict(self, iris_code=None):
"""
- Predict environment variables for the given iris. The environment variable to predict is stored in the dataset as "env" variable
+ Predict one EV for the given iris. The EV to predict is stored in the dataset as "env" variable.
"""
iris_object = model.get_iris_from_code(iris_code)
iris_area = area(model.get_coords_from_code(iris_code)) / 1000000
diff --git a/predihood/classes/MethodSelection.py b/predihood/classes/MethodSelection.py
index 71aa5f25..dcf8c37c 100644
--- a/predihood/classes/MethodSelection.py
+++ b/predihood/classes/MethodSelection.py
@@ -6,10 +6,12 @@ import seaborn as sns
from predihood.classes.Method import Method
from predihood.config import RANDOM_STATE, TITLES
-from predihood.utility_functions import draw_features_importance
class MethodSelection(Method):
+ """
+ This class represents a method for selection a subset of indicators among all INSEE indicators.
+ """
def __init__(self, name, dataset, classifier=None, transform=False, parameters=None):
"""
Constructor of the MethodSelection class. Initialize attributes.
@@ -31,33 +33,23 @@ class MethodSelection(Method):
else:
self.classifier.fit(self.dataset.X_train, self.dataset.Y_train)
- def results(self):
+ def compute_selection(self):
"""
Get results of the classifier according to its name.
"""
if self.name == "feature importance ET" or self.name == "feature importance RF":
- if not self.classifier: self.fit() # TODO: a modifier
-
importance = self.classifier.feature_importances_
- indicators_importance = {self.dataset.indicators[i]: importance[i] for i in range(len(importance))} # create a dict to associate each indicator with its importance
- indicators_importance = {k: v for k, v in sorted(indicators_importance.items(), key=lambda item: item[1], reverse=True)} # order dict by value in descending order
-
- if self.parameters["top_k"] == "MAX":
- k_best = [[key, value] for key, value in indicators_importance.items()]
- else:
- k_best = [[key, value] for key, value in indicators_importance.items()][:self.parameters["top_k"]]
+ indicators_importance = {self.dataset.indicators[i]: importance[i] for i in range(len(importance))} # create a dictionary to associate each indicator with its importance
+ indicators_importance = {k: v for k, v in sorted(indicators_importance.items(), key=lambda item: item[1], reverse=True)} # order dictionary by value in descending order
+ k_best = [[key, value] for key, value in indicators_importance.items()][:self.parameters["top_k"]]
self.best_indicators = k_best
- self.parameters = {
- "importance": importance,
- "k_best": k_best,
- }
- elif self.name == "heat map EV-agnostic":
+ elif self.name == "heat map":
fig, ax = plt.subplots()
ax.xaxis.tick_top()
# get indicators that are fully correlated (i.e. corr=1)
- temp_data = self.dataset.data[self.dataset.indicators][:] # get only INSEE indicators (!= CODE, AREA, EVs)
- corr_matrix = temp_data.corr(method="spearman").abs()
+ temp_data = self.dataset.data[self.dataset.indicators][:] # get only INSEE indicators (!= CODE, AREA, EV)
+ corr_matrix = temp_data.corr(method=self.parameters["method"]).abs()
sns.heatmap(corr_matrix)
upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))
@@ -70,8 +62,5 @@ class MethodSelection(Method):
if TITLES: plt.title("Correlation matrix: filtering = " + self.dataset.filtering + ", normalization = " + self.dataset.normalization)
plt.show()
- self.parameters = {
- "fig": fig
- }
else:
- raise Exception("Unknown name. Choose among [\"feature importance\", \"heat map\"].")
+ raise Exception("Unknown name. Choose among [\"feature importance ET\", \"feature importance RF\", \"heat map\"].")
diff --git a/predihood/cleaning.py b/predihood/cleaning.py
index 07d6b916..63ffbab8 100644
--- a/predihood/cleaning.py
+++ b/predihood/cleaning.py
@@ -10,6 +10,9 @@ log = logging.getLogger(__name__)
def clean():
+ """
+ Clean Home in Love data and generate some charts to study distribution of data.
+ """
# 1. read data from Excel file
data = pd.read_excel(FILE_DATA_HIL)
@@ -23,7 +26,7 @@ def clean():
data.columns = columns
data.head()
- # 3. clean data
+ # 3. clean data by changing misspelled values
cleaning = MethodCleaning("cleaning", data)
cleaning.clean()
log.info("Many plots have be generated in " + FOLDER_DISTRIBUTION)
@@ -31,25 +34,25 @@ def clean():
# 4. distribution between women and men
women_men = data["sex"].value_counts()
labels = "Women", "Men"
- sizes = [women_men["Femme"], women_men["Homme"]] # getting number of women and number of men
+ number_men_women = [women_men["Femme"], women_men["Homme"]] # getting number of women and number of men
colors = ["salmon", "lightblue"]
- plt.pie(sizes, labels=labels, colors=colors, autopct='%i%%', shadow=True, startangle=90)
+ plt.pie(number_men_women, labels=labels, colors=colors, autopct='%i%%', shadow=True, startangle=90)
plt.axis("equal")
plt.title("Distribution of gender")
plt.show()
# 5. distribution between ages
data_temp = data
- data_temp = data_temp.dropna()
+ data_temp = data_temp.dropna() # remove NaN values
ages_plot = []
total_plot = []
- min_age = int(min(data_temp["age"]))
- max_age = int(max(data_temp["age"]))
+ min_age, max_age = int(min(data_temp["age"])), int(max(data_temp["age"]))
for counter in range(min_age, max_age + 1):
total_plot.append(data_temp.loc[data_temp.age == float(counter), "age"].count())
ages_plot.append(counter)
mean = np.average(ages_plot, weights=total_plot)
+ # First view: bar chart
plt.bar(ages_plot, total_plot)
plt.axvline(x=mean, color="red") # draw median age as a line
plt.xlabel("Age (in years)")
@@ -57,6 +60,7 @@ def clean():
plt.title("Distribution of age")
plt.show()
+ # Second view: histogram
ages = data_temp["age"]
plt.hist(ages, facecolor="gray", align="mid")
plt.xlabel("Age (in years)")
@@ -82,7 +86,7 @@ def clean():
plt.show()
# 8. distribution between geographic positions
- geo = pd.concat([data[OLD_PREFIX + "morphological_position"], data[NEW_PREFIX + "morphological_position"]], ignore_index=True)
+ geo = pd.concat([data[OLD_PREFIX + "geographical_position"], data[NEW_PREFIX + "geographical_position"]], ignore_index=True)
split_geo = [geo[i].split()[0] if not isinstance(geo[i], float) else "" for i in range(len(geo))]
set_geo = set(split_geo)
uniques = [split_geo.count(elem) for elem in set_geo]
diff --git a/predihood/generated_files/distribution-plots/distribution_geo.png b/predihood/generated_files/distribution-plots/distribution_geo.png
index 3cbb5edd82f645c6ca24b0c88992a9064779a630..5abe54d063395825b32ef78a4be13f790738b3d1 100644
GIT binary patch
literal 28987
zcmdSBcQ}{*|2KXjlxQL?Ng7l}5)mzlN;YLj2pQR%hK59xw!L@C-VIU7ipW-q$lm*Y
zyt>Bc^Zgz7eILi~IDY^9uH(9{!rS|Op0DwIKGy5JrFi-{3)4C#ilSKLWTllUiq?yw
zXxbMu;5#DK&7JsTk?j#V)y4S7ZSnb=_<D)8>{(liT7HiFM-z82))e0qvpaUiPQ}XD
z&hea$5oK}C&ibO2-9=LaE(aqUTT?4bVSXWg;ayx8?Ch+?1qA-{bABru69F$uNSdO!
zC^_lFs!q>-x464ij4doI%mf50U5p7n!(v$e)h6&+7mvxU9bNI>E)ut%U6URb@(B-o
z8F+tRm-u=<_IO#_RWJ7>%l(WxqhicqocVp`M(FqY>SwN&5<jzb1Rj33xpY7;_4|}t
zzPPia=Rv*Ht`Va&&17{NiYhj3R;r@MEpU0Rr114KW||fF?)w%0pMH_eYDcq>VfuMn
z!TfY@V~%aCAE#tOclQO!@diVOE}m`MdZHid*1Ws3KIq=PNVB?xUy=S&%8&OL@yI=X
z{CM4do2Uo7wPIeHR3;p|&F<0_>}WSoFrn+XasU4P;x?T(ebxx@-!3s2U5D$p^sY0j
zi&u$dU9loeqCb0R>cEYqUTfE`O;C!wK0W#?vLaZbOMHH2peXUu_jfMiUrj{Ly<X((
z;=(5+q{PX|nQ42SWqnNL!3FJ)4+S*hm816>yj`4O)ut=q*smJEeJmx>Z_9DN9fE>Y
z4fgo7M-G#N9qXU8=e{+_&`fq#^k2z-duu(jzCWjAXHO5gTAgvxU%!404-HjE9J^gx
z;OZ#1;Q2`FrSYA{bW>X$!`Rr^j~_ptys><f`mI%4E$0j{6U_Uahofle=<L6}U4koX
z9URzb$uMYUV`DS=<Ds5eZJd(X=a@SVeT~s!VO)3C3cq7t<=EdGpOmyK!?LAuJkQIE
zQcKm<NlQ!9VD`fJeJC$acRy&~bDt+(`qrvNhbXH3uI0E;#*vK0JR%}$u?nGI?(?dQ
z4!&7#-jZwYGCNd?0o-oUoUIgiKQNG6R8%x**V&r>EDsNl6LNAn25l=>t{fg6O?S_p
zTZ#eK@aCg_c6Q1JiaY#{TchiASj7BGZFhBKU7@>c!AyUi#$0lpS%KLJPGMu;n$Mpt
zLf_n2zCL*)7MT)j0_8N(;`p(o<d|BbdWLzba<tE~jp96V!xIzrQzKnL4;~mho;EW}
zn4O(XX~@dU%j<Ix$a1Hliu;ljCF#0<{)}PdbUGv_C+E=TI?=E8^@Y(^xrsK<B5fU=
z439Y%``$0|fp(+aN?Un&E?~f?bcHD@YfEBX!qPSTJ0v9X!cOMwX5&?s=I7^k&SVRI
zLtp$kR>8cj&^>eD{rmSm{{9cMva&wh=Z##W=cd$>=csXSn|y7)vu(-Sx7{O^2kRZ%
z+_%eNENWj?N+qt`A`_<?uRJ_H{-rEn3){MN30Sr4RcbVp^^BP3d_Gz8!X_20cDz{`
z1|>f6cE4(DY;A>Gv))+HjkGwn#XFDu%xQET>xnKala;HvLcf%C<;w7b^OHw6Z{GYw
z@Z1#)@CAv9<{Vp>p^`NjcE6PK?u1#LKUezx-u55Ae#MWD+F3HvQG*M(^lGc3xft$a
zK4qn)$1N=_^95fC;;u)!DtjB!3_ny<sF(P#*;<UdCGIcjpbd-UqG+kRO*yt*_n$pW
zdSUc2;Y^wVy-tNiU-@HUai_tsD_5@k)^*z4;&Pf-^)?oemtpqfdjF1hoj>#p=lih{
z#MaVs;oJ8vWvdlaRG2lN==4XojRVE^9Yn|S8XC?>PPSjgz78@lFksfOw<vI(_;PFY
z_H>&cAKW(0tTINi@lrYaT>mn|a?YerPeq$_c{(mG_9Er!{Q2|Us0KB*m6EaJeN9c7
zk3Qn|3MN{g_SU_)bm`KIGrJm#Jd1ctvAdaASn7fuTjPt153y_zQ^d;74BoVMEw6xp
zfcsdTMgo>se{L_o<m})ZahK6J9^L~7w3U>UG&9VCcCp{azZL@DXQin{6m{AV1H`ms
ziIS}B%@CI!UiZn48!t4IbHDpZ)J7g(zWeNpxBauxCm$vpzRJL-SC}7ld#A0*rzg?Z
zt}Qxt?AWiTR)y7ZN|C=kY=4%oEf{*YUNikdz&?w{sG~QQnRb>11V#mHIi8@EWtpky
zgyku6u|d$2e_qdt;knlwiW9xgm0%_|zUG>NoRIvIv4-z1!|e5j{=GR^vs!1*p2d#N
zj~<d2$F{c*mGGQ*vma>H>8c1;d1?H~ZhF)-Stn0@x#Z~GT)STpzdpzA<mHt!`uITY
z^l3j*N0lOFzP{rS)b~nGt%*Kq-kg1Dow!|Gj6x{C+=mYzl-1QAw7O5|G_^%`g?j3)
z-hTQ%Pm6uM4xeUv+#X@!9|HqcJp!*-3>nZS#Jn+)W~F``dm1wmdER3i2TM4rtE!^X
z7=L<#;j2Er$(=`T8n<NsyFtjlzxnHzQdIf66dj2m0X<zQDXHnUh51;|`KjZm(V4bB
zY&(tDRGoatpIlp8JDdHsSLV-TY~Hg+c@4j|qBj$pUA>Ouuavz0n%Sw546{1E=;&y<
zK$p?(b4$H;swXKXswdl_qMQs|#(5yFl+*PT>Pj6JH0%2HiED+8Wpl+w9B(BiCbnd=
z-#+Fy{!N4|+w4A_Z0p1L)H>9`bc?2aO(rwreS9i0z7GWS_~m+93-VU2TE)6%O)Od#
zhPvg;ac%8q{My-JXfz4B1uobw5f+z9?^xT~syH}grkUrhJwUe@wb9l5OOleN<`b8R
z{@n@+3VDY92j{aRA|ghcZ7buob8HgyiV8zq`}dR2_m`TE>&qYh@v--{_fpd2YB5R{
zUtb)<t}W`&{1CJ=HnngX6<Cf>E#dwFNlDvJ>wb!mty63vz{ht26RwytxLwX)a{Q6U
z^snuGjTt9vN}7rfF3i3pr5XLFu_6&|Fy?e{f@qtY3MrJh!4wTW_p$5dznZdAj~qFo
z8hd)<)Z}FMjZGfeQ%m#Y*b^gL?mMj5t`Kq`y$bbPEkgR%f%P}kjGS-xrj?F$?oKJ?
ztiPAVWWzRy`k|V6@oY|BUYjz@!q1M5OP`Ah0@RA9eyFE(1=@C1@b2G#_VD4uIn#M{
z&t5bCDHxkkdH$^)|8KtF7$nVbYWN4e&O7@{yxn*k%90TkU+p&g?#{lxWHb_1yZsu+
z$O@8iIyLmMjpof~s~!`r%!^7NAMCD+IJPudx1jONeyl<tKR@+{{5q<M>bvO~7}oVR
z?X{$#qo~*y)*VMACz@@2?%b&hKYGJ5KK1wSi$6L$8wxez6c6V)^x1R|ZYQ|OYxVoG
zvZ~lnj|7jo>3EB#OwD|!w7!NI`KPMs7nI0;#FYg-eE5*lZSXL<QA5R9RD{{NxeVh^
z>*^BJPKJbpWUk1;$JHgBRngVeC7TDu*C<1Ct6W!QXg#_O*{vVnygB;gLy(HJwDi30
zY!N|eUY{pB?<uIJ85~nfP?b|SFm#QbU;DyR(}(%9!ykQa-mLoh>FJN&-b8+#+#Nf2
zesq|{+q%q+)sIz34D8_NKaJbhMCILf>QNX<GM|)`u1f6bI`h<m7<Ay1f!N5=b%`3i
z1d%B#2eJ$3CZTHo=!$OYSbGRiVlVfWEn{z2E64_FW?3ft`ZBxE^kunphu^r^MJ~Xs
zI>MXBbbfY<Y1y(b32KS8(I?l)ks4!M@hl$bpmO}_&6|tmgT?DICnNF3)0ov!)Qd#D
zA`fx9?z?>INpYwc9}UvtE`EJ6W&i$R30XMSOz8J@ej6ofGt4#Q0&m~GjbfM-7q=sC
zq^B;?WacIMHnvGrR1}ZaQ~Yz}=uwL{|BVtcq@A8ScaBsuANJjRw@V_@3`@-x=H~*V
zS_@ov;_lIG*X=of^GmW8rpraXoGj4T6>HbV12e=O@n*{0msrQAp`n4wqcfJpe!FD6
z>m<!U;K|#E(A;X%O;zLqZ``=iGGJbxlHA#Op3J?T=UhgveQyk^Z_hmishl6ZxogS$
zdNs9rFI|(4PA>m=&#|UVi<j=1R&6PJj6O72wCB6eI5dTRuBjPqvMK@+J)!Ztsj^am
zhK6Q(yvZt6^C$o#0hFIY59*=x{aJk9Oa0rXOIj8I^FG~YQG+>402(~Ii#=I8XV|A}
ze4tGabx+v1{JFWp$&)H)&OAaLA=u&hbHQM-OWdg^87XR+*{3HWC-CEe)&jL;O_99H
z@BRlZ9oLI-b02B*m`NW0{WX%Ta$ueM=7y;t4PCNQMQneb>TAi<$a5U{!nnyTj8`Q_
zHn7#DclU0+LhYjY>BRB=mPC}lGxPKF!^6WLKKB-``VTb!S^n6zZbUsro9rE4E-tS6
zv=B*GvKWPggnVw@QgU<4fBpLPaKd>kxpq=PtBx*z|M|0UeHy9q2j@ob0T6GMb3vOI
zvFVgWM^!xT&sn!U)C~YOD{;InR$<epU<nQE{LggD4}@W7nsvPP#x@K1)83wl+b8tY
z#zzu7$+&Do)e{kOyV8yHt4lWYuPx5BXiC6!X-IZ~8Pt-kNGdT3>`RFUj~>N6dc<MT
zl3O>vq6<x_F6=Pf)~#E=&dZ=6y{x!1-y7pO(`NGX)XeB-g;b{`^<+(P*YPBP0UkNC
zhUei9{msr_ZQs0pogx-bpv72s^-<;iUQ&*y$NOA`qq~;ecuJH1;Mubc-rnB7O4jJ9
zqVL&kY<ppJ8<W?kK3JQor-$-^8uh7C8CYyX;G0jB?#J4Al^@@}MQlELU3=W=#0eia
zcX!*2ndxc1ty_Qi%o!US^9c)o9-r%ce=kdS7qA;0pDK+tK*W>%wlUwnef#}DZ=wE&
z-`V6ZiSyeq^Q*U=5>(z!Q9In8>@|%57h|_uI~AeWLOYM$8ua8z0#<Dq5CEVwL5woK
z9OqnveTs@O^*bko7Xu{R3k_{!oB3}V&{b!IL1K+NB0|T(&Q7*RV+L5F3^r)b=U9K9
zc%Z8j3JUeWeQH^jI&w9bi2ETSDd1?jX;zMoj%YPqIztrIEHghdul?xZL(e-K`Kb^o
zcTF_#m+lIuPpf0$<QEjE#hh9j#HaS5-wiu=YG$TMeL_86`PlM<v(GWgi9p_RfzxBX
z1Yy3U<)FN@>`a?T$NDk#LC3@qAWm-0OFz?PYbeLx4dE!LO@+%hNofbxCTa*AlabNY
z*mUy2uIjs+kJjeeo69NRShg`GClzHXdmjo6V0brBcObu|qfLg%?~yk}Yc91eOHxm)
zc(y;XwpP{I**RV|fcsskbvD(IboS-?C&LG<+YhDev$C?<t()JVuZA_h3U_OSITI17
zI}@6<=VMS%kWtbyl~d)7`1SW~ZEc4{a`tPJUg!78Agv5-LOoFp{8(msmr6{z{K9xu
z_)!@%!C&er$8$NjMht!qtoj3ecpq9aKDFLix%#J(V(wYLdAIZ7jSNKo$nv9rVA{_U
z6i%JK{{QDMHt<uHjp=d(VA0ajo=-ai-f^NS4TQO1wC3bLqydFea&Gi95fpSbnVOn1
zEnRvKkOzB=4QmrPxN{;JLm|fVKdRQf-iF7*CIlNoPvAv64Rs%XOiCT#M?-M^OB02X
z($byVw#kq>=I^g@=FAxn9ds10qqe{-^NA=cW=)xWM*d*>W^>bH4i?y2>fqt_-Bp`|
zcvWO>-MWSPSB-WAQZFe-aPQwGhL?wjr!o~AzBpiiGjsbL?39Nk1_9_=Jf<jfgra*N
zlp_$6+{60yt|cA-@b<%<a(T<~OMP*QjKx^urlZ}}JzrnOp18kbvmD@MKx(R0imki5
zdr)AYO-Ug)H}}O#W)~<Vf`WpTJy;q$WWD!M#zsc*D6k*E#Uuvu4IRr3(mKW4%rbN{
z=K_Tc-|di;Y#HD0Hu-$$38_ez-$>FqNxpdTqBdDebnEu*y&y<wpXTKP|I~-xRPCbu
z4!<A$VfnFO`OPgYBT>2-?BmMa#*~0}5!#MCX`&~B_>3I$|5Ft#Fnks&E-tz2_Fa6`
z_Jcsl4XBm}+^5sAlTBtqKxZc>ot#}=cXDwZ(#f?`0&p*!{I=A{#H1;tcVvA00l-0m
zYJ3Fn?iu%)@w$BH5gs|rg9-}BSsG@F;$md>l5qH~IP&w8MH_S)@PWF-hp*qf$#5An
zlT${k?DLrD%W6wF`_ectYU{}d7s~E#Hfzbv;E4y>d)T5u<g~Tenisz$Jx)^ZW%JkT
zYuMV_+DZ`O`xm%L%k|#<1WHlO>;59q|2r7Gg+>sI=&7hxrn@HqmfA~W`LW+$Rj^LC
zUB|C_ZI+>Jzb#N0cW=AGi9m;e)?}y=-2ulrtLNutEM_LjDiPxN$|wi~4oVcn<39Z>
z)~v0tV4FfnQg*hO<Z#)Rf8NNLYH4n^BQ>w4UeD9QmF@<>l!Y<wgYtV-#;pJS-5n9P
zNox_aTG^&7t9o!Y9yx->z{$-GP%|hFgYyj5C~h{FE>)ueM}QaffaL`6{Nbf2xwkgs
zRr)sr))u2_bmOwi>gK;_Sv~IN#P=E2HZcx@o7C@2&5zg~;&dcFJ|4AmIraso%2B+a
znA$>Pjf)@~7Y$+bt4Zju(O;hfqe4CB&jJC73ZV*8%tEUuEYRe7U03<u`qYP*=i~y^
z`9}23e?FhecAc=4oS*!j)9CKzHj3u{h=cHm>l{nGfWe^=Tu2$$$#eLsz6fjp1GvOX
z(sjJ{q(DK`{rjwdRiFZdDOEx58AE%2`tNT;CnEZ#U8UqiH?A!~Gb4etO*F0^;3c8I
zukgk4FmTZayUs=cd7IUz=wLf3C2M97Fu%Hmd?eQvn!!Lw0R(t%%fo$nos*%}2gd~J
z&N%mmAvz|8Z~OMr{z(uzLjw+6JBL0Gh1#PU5D<WO3{KEUCE93&7Jk%8gMkBMnuee<
zTDHaiQTH-@*SV~WPah&5JtO1OZX$&{m&;Ncgbc0$q4OrznMFvDU){VV%S+U%H3_<s
zI*_dzFbEV*GsjsdU&2DzCls^bDykqqpMQp5A76=a;>zDFKe)G=1z!zc4ln^#p%k4k
zgr0L_4}s;N&~mL=vj+0p+kV_CReG+I-5QXS6uvv_MNdOrG2U_kKVUSz$VL`KDlplT
ze=d1<J47Bs*Z)mn&-;$f*U{0zw0N;1L|5=B;l8ggOZ%Utr|$)V0%%kLT0(XIj6wIU
z1>jX|Si8gvptuUErG0zxWv%|5Tecj<`mTDm-<IFi4a(sYDfe7p)W}Vqvq}K4M9MlH
zEN&zbYf>3P%KE8OzCi!1n>M9BfBsx61^3+UppW~R=x?#dPErF~grM{>W98`3_tMn7
z)V~qsLrLQdO%cO)%y%|Qs0WMN2l7BYdJmKjF+vTbB<NpssW#SAcaMi)eU^256+y(n
z)h&nlFs7EvvGsk|3ZJsGm51_&3a-@qCE3EFzrWwCF+JYE(6Fw^(~}4ZgkwS=80)Qn
zh{lE=Eb+n;`-Ph?a2eZE{YfM`DoQpGRrq(~-0ZB~VEf^w{e^{vIUc>ez2-A>KhFOb
zX54lHoxoa5hBk}`wbUG(^u(!C_e%_vKxPS5KJb-su&Am0OtN}j=0}U8BW<?xJO4(J
z8a!CWSNE)<Ci<FUk8?UpL;i`y?VQih%y1mg1p)8*C{UydvF8-?7Ee!4xiuzJCm!-!
zm@CN3XMAFso|)-QEt>x@)?`@f_v6X_*(~&%3((;+2cS}@>g$IE<YHJh+=B9O^g6Q=
z5sfAX&y|#vm|}+!P6;)0@A<M7tB=YC;;)?p1F0C2rpAiUgQ-|}3FqEi4~$}9S%)5`
zg94Wo{1TE-bZl%>Ma~AGu%I1EugQY7me{k0F8k7V8DM`k)bq?ta5Phh#Zc(0S;p+4
z!}T>~RgO!d<PHrDK^9g=Mc50ePd?<pS+<?(s#p;s&co;N*Hb7f_j%O*Pl%Ml#y@nO
zHVmz?Mau2rGvHf@x}vr9gaAX1ZkjlmW%Rdjd}p1=2`tLG;>&c9&tqK1dQ5>P_;mBL
z`$Yc+sTUKyCG5JbOWf%(PS5THPz{jkF$4;m+y*cAd`aH<>zhSA4IM)q8klIU21Lb_
zY4w#rvV()x-ZyUO4Q==fnND7R2j|F|T_%L)nTYuVKGnD|3>(3oa8fV|0avbGCA4t$
zQS9R46<dT!Q6*w4dUNdCx6)^m)xj^-UcGv?Ef?F+>+rjAn)^}h?fNK1JE3%fN~%LV
zJMmzbsSW8X&D<zLH{j1Kmzq3OQ^FDsj+@Y!VZKvPP_fN@BKY63KKXuoXq?QQb%9aD
z$(flLsK<gK^J@CLA?=?wN88@!xnSXXvJ_>7z!IXKR7%Yxp?O;T-NtyR>8Yt-5A{5h
zP;m*ne_?b|t_FB-xKpPX7nJQR!+sg%{WwNg<yTF#QQDc>1ht)VYbI!uv~ynkdp#VS
zoGDdRiW|l4WYJB6FpWF}@voC~N&o9LWxE-R|MG7%fiqOV!utHM;d|Bd$he%#nq?-9
z0wN^9>#15SKCc7D=U0|}Ryb(U#Q3K%e<~C9B03SAEwY*uQJr!}DkS>+6hft1-U{Mk
z1&d{=Q?jy(psia;zs`_&mwXQ2FE!wyJOC-SkzI30N^qFDeJjAvXUKxl=(#5X!BpN8
zjbGm@fSSaB#=+JCWD4YT?p(Q1(nU3Jv}j>YVT+8f;J=MmkxCpWhTqyqxZqh!d$
z{Xnt*^gVn8<dD%i4ZUBjt^@FWpr1mpz3?x35ZtomjpfgQz@mlurg2R%T%{Pp7MkaB
z_qJ(fTZi}Mq0T_e*7cmT7qMz>eoIC`@NL=dcHk8qAhOTc7k2Zr&SWpRj{mMk%^gE8
z`lq=+rb$9S_L!eCsY}+<K)+R%k-08z-xGz&i6@QrZ$xEGZ3VN^Bh0rBeib!*GmUzL
zJw=uuU^$!z;wZp(*n<7=4>UoWv}}%pGl%Oq33w*%F`G>^G<*k^TJ=BI%G@4y_$pa;
zM3gw@!y53XjH3Ch+k4?ICi)aJ>4|i{<rLx9>(GvO&&nyof!qt|MLdFkE|OLs6N*0Z
zC{Yp?CcL-3f6s!=QYRo$BJGpy26B5cFw?RnA*;<pOC{!1H74?+#6;mtzpxPMF=ZKu
z$^9eaVc*z)HH1urHj4j${rYv=b=03?c~tCs_wLEtk>%$GlmWW_MOqZUTgHfOfIq_q
zci7Pk67~8=kOnI&YZN5s1h=W-Ct{Z(;9g$rLN`GDAW$4j2oUyDIw}=y(mK5RCDvR@
z%)UpJIEGJzP5f4^UM*tLxR?0FPb6J4`<|J8jv*Eg`I*P;&~b41L;zXLnktPtU`#s)
zlMdAkY)7PSNI^th@2QEtO+?9!5)KglXGH(gXO*$J*x1;jVaHv-zNOce#L|By%uXph
z6AK5_&@+7JOIFQRIXFzy5PalWHp`hcXQ%DE*bs)rd?Ju2>=v#0XZ;Th--P)SsG_V)
zxPOxsxJw;Gz>;_Gbko|(%TEO#aIX2gUA$nLgr%YmZbl?GI>u#jR&9l(FWA|o6Tbpl
z;IwcZQV4%#0iJNk1cILm@S+a|?bJ}m<lhetPR1tqI?jCxM2%-^YU=A;1>geXa?X5%
zBKWezCQK8@X5A13blvghMxx_D>on*|XGh05bE~GzWC%~^4HiE|)44G8twb}|E*@eJ
zE5<^!)#)hAFo!PS+Pr~>`yvYe2a&7((^S3jt^@<2N`0*J0J?23?mXoTJ5mH#nMhZG
ztnfXrGA`#svj?gm1TX5g@MVCx%sGA$5s~_&_9Cb@A6r^lE+k{k_?4A0huyeQY83kT
z<cSl3{{H^AHeV#dh(!N$fjkt+Obh-0YSMn82R(Jyi(0C_xz2Lu3$VH(2jB}Pxc2R>
z)zCz?{`%HePfRvKYMXOj;>|eSN0sEC?3L5=n2KJuK`aa&j(n(;4sl{Eev_Us-1(m9
z|3{8oflZT4oLQ_&8FLKc4qxA9T95B{L<w3htFPC<57a)#$b%V2pwnHbiIUrP{rO|8
zPe|0Um+2TG&!|D5?P-JTHLx&W=}-X|&~E(qSz<?`7yyljfXT;xE4jTs8ND-15S!}c
z{T<=(JP9Wz^b|WqHP68!<KovE=?Mp98&Y6gU3llaR^vdaLD~*jmH@Z!O3@g0{PtZ`
ze3H|pd$T0u(7)+{=tc@gq<u@*vbB3Ujc?#Cyw2f!ZBSzK2B5-yM*Hgg@dE?(^f0K~
zHa$H(S}weZF6Wn8G4Nw%Wt7}KRnW_=NM%e8eJAw`n^6syR50C>xUCnLUA%kG3zl+V
zpS`Z*7rn1tYiI;~8wG>=^<U&>6@?UpSrgr0+l!Yk$55Fc|I3bcYo;s0WX~`xT?Pyw
zS3_V-k>|pJuS@`ylxU?ar42`-Kc|b66lj168%IYCtR~f5JJW)N`I&!e-#E=Ef)(iB
zL?Y)!`9sM<8&`@x5p>E98V7aN*K&~#aYmBPRG)u{ZwQu`?bgSitelGu;g9V{DU>$r
zJ%4>+xQp^wn456viI1&?iaR&e6&e_YN@a@G9`mOs&;V)^lLwn9?m~40u*XA4DqH1Z
zPY2gPNR?;Ca8pMyD};3p6wajltPB-^;;7|{^~$zpjmf{jV%yfO@z}{Zjb!#AF8$M+
zmUv;Sd?|FFvB|JPdHM12;eY5XfJ`c&HS{72^d2Hq5$Xsh*$jQ-pE1#;7d5Nh0k}#;
zO_(Cjz&p7CTkmYxuZ)_X|2HT}bEHwgrT~Pl6DXREfVxbem}86m!}veFo*)t!9t6w*
zyu#$ng*t8+!H@gv4AAmmsDji>JY@i*TQ2zK-BRCOw9csedyUIEHXih7mUhH<%Z~ny
z4M_{~@}h!vYpFZ7I>)wL+IJOGZM^~GY3MI$9I>9fFCFOV?BvitKqQmJm#42G?dX46
zZGX{%$Gtzi7!trBc5u~dg#+Q6JSGnVSO!L67~`;^H8U?hMtny8?+%(o>+$ok$^D?9
zXh2R}nRcoFCJh6F2*8>U_wlp;0jGe21t3Lit5zA9FNOW~`|HcPIHfKBd4bV~`%3ES
z>M7$Gm~IFf|G=45RIL2dm{(UB&9T3b5(wnyU%VE5x7F!8%j%;IQ|KRKwJHk4Fd=Bj
z>`PKSeDX#=t^Z9s<os@rp<tB}bzho&{(T_kIdg!-mryvgwfP;KJ6_Z0WM_9jHVF~R
zRnP}IUMFG}1wq;zC}4POtZtxa!BWeOTZcIcNfwRs=c9-VMph$%IiQWGCzeU<*zd2q
zW9a_S#xIvzh$9R-O+pTtnK0qL&R_h$($bN<^SH<xHlJypYozLK>&>=I)JRna8YgZX
z2@b(B^=DYNAr9;KL}1{(d!%5Z?lpOQ@9$56^}@Gr-vgp3KMSe86Q-h~Li__P!Mt(e
z2sUsK#EzH?QJD(hAvPAe=Ur!BOI9&aXo*Jo|7?B##+(qHN_#^3->m>D+qUa88vV!2
ziMAp%HSpA@XsP!RF#~M4fU+ML1vX+@86x>a!Z9VpW7;(D(!wXRjGtyRS;UllB>3t~
zFmBZELm6MYl>2lX2p(~QA*lsDdv<Z=cUw{0+ZFq;Vb#HU`x=Rkh~6U?2w!*W`+cjO
z<jl;>QqI2eT>n3!Pgv{4N*s@APMtpe#U#`{283sIi;h~P%$@u9@1L*F1P3PjBoGLS
z2lO{tuSgeaOCV41U#+he#!hNO>fbVHiZw2f@!WZa=O|Rr^g>VB*~<&$4_+Zy55XW3
zW+F`<I9n3Rfox5QRuO8^XH+Y;jT@7ZR~hSX(E<}y!7z+O%i!IP8=q$woSSIF)E3;o
zdp8Vv=f}Fbx_&3v8ersdF71~YS@FAsc7J<&i;0P;2J*uwOz6gnwIXKsD?%g_NiW>J
zTOOnT8A!i@4hS6tR{!F~q<s*PGR*6LiW|GEyvakNbzsRw-!A=2FV}Be^S6332t$Px
z+I!@2|34y^q4+Fjw*3n<q=nfLe_f{z@5GcB=to|L(4Hg!;Gs@^k=DRHlo_J%B!Q%$
zZ$_fI?%ca~Z#_5O@A@MS6dsC!LyO)GQ4&+=h>h8oB3p!RmmK-i3%K6xyMk97>d^08
zTkQE#GdZV~k;q%6=XOzud?<#8hyQIJCt$HEAZ)Ow4(JhjBLFu5>N-)&XCa>R&O{zX
zJ32T!c$I;fBMeMNi%iR{b9gObQetSG1p6x-oJ&tmMjcL2PksVfnP@W9GD2o>+k~^`
zG_K!wF%Dlm`L}<9F%u2XCIiIj2$0-=LH^g*mXN#($r=I3zc)y`facZs{M;L(tDEu0
zOnv}x$UvH-hvee=GvI`{BL(^A?=YN+82V2q0sZ*}qEQd9B5zN4=m7(_KZ2ZCrq&ah
zd_<wZMpIn;N(`!jTrG|MsE>9hX#~Z^r+rFMNkwprx=Zn!?fRCylmik#8cl?VW8wTs
z&%p|hC*c$I>$oznbMHcE4xpzIKaPwn3H-pkgV(DH^X@Nt<)yXE3vo)9U(pX)H*83P
zYH|UKd3tuFQmw$ne)SlZNAYv5T>1l`4#*U!L8?PH_-bUe1f-=nty!7c0L~5LKLMK#
zQX4NHU(bs!G5}e7{%DxF(bOb_Ts_<M<Zi?sXfb&~NWK06UmWu2kmtgmr~mI>SRm=i
zgn(f9N$8Yhu%Ie62Cs%k6dU4l0t*N^hzBm?zvJP)uql>M)+3gt_`GbLNl077;3YXm
zo#tYXVlNsZmyi-pH2RTp;dO{@5oD<V{Re{Ec>z_*dlk)!vh76LK|W-xbd#rSAPG&N
z9VLP@1xCR~Hv^w{D1!VI3{Cg(K!$iE`)<t!_f24x5BPNw;E)jUgJ{;+@_Y4*siyr9
z^ged+(G)Y&a@|!(5qZNShxlVX*3reHp~1l&cQu1`(E7im=p=$e#lh5`2DQP+9gN3P
zCUG!m&`}68nZLccG68{-!w;8IqDfT;={Q!d+zdf_V$&y(BNomBXON-H>dzlJ4c+(E
zzFu0^`?$6^m?@|$%8H72VI+qGlk-YRwK|-Jp8Mnbcf%Q00o{CRr-Ky2P`Q9vaKzP!
zWQ_4tLn$K)vRn<ER1%NaedhTgRA-WP0Qy&aevGbIen&0MAO@<|b)X=Ybet6ThUe$N
z0lwgZGn_|^gCVe?Drq=4IMApuHGHh7_<{iFxu@rH&*QB~{Ud2vqGuvv$Y0DCtbhwu
zPt{FUkeZSv4gqmuh-ZZY5J*gY`(HK3@SC5Br9l#22vqR5C(yLNV{6nZCIMWs6%e=S
z&8XLXnE`RO_v&(~Dv}ufqc~v)vT<<4Vee>KG^X>$1=1j45Dyt1K|pu~B^QNH1=fZi
zWrTG40fA3q#;mu>4~Ju>21ZA(oiT!B+Dpm$HQ;8L#+^uu&CN_a<blrC6s#A1gdQ@^
z-lC6-ytLFWm{5!?>!PrLzaDAbN>b0r>VQOBm@kE%PG~bR66gOl67?H{j2}haet8mF
z3)*8)Eu_RxkA;tux(6A9Dch_I876y(F)bMF;gJy)8=Ds`LkPyuFIvQ^XphCu^n<w_
zg93OQ3u-A;p9XIwCnqN!@W-L=pU`pW-{f{k^Rj1N4fE$BlQe;NR7@m&?IBbXZ!~nL
zy_8k{(Df%!1+f)~hOg%~<PE7khJGn~nL%2&=fX@V$sHj}mk2A7%s6_vh)Kn>EBqTN
z%QN9iz2NYDfog>$b?$o2&88WeskwUfkAw_~?9wy=cMUe^-~%)IlpnZ=AG1i}ATtTt
zsvxED;~AtHD}uA(SP{XMPa`D>pBx8gI4}zMVTZ7=3fW}vE6I<LR%%MHq;{Ml$^|m<
z$|@@MpdU?*;7cScPlgfIF^K5W@GkPs79%WxV9TJ-T<#wREvaCB!#1yx`7`!9aU&uA
zoPq&~DVQP2VN{vMxtD(V!oaQOA;Q?!dF~WV*mFKR8Uhv?3tY2wv_hraRdZ~N{utrR
z%=e)(eo56!#ofyVlDsnTf&9~b$y_HGcy!#a|ANm20HlV5=bk-#2pfa^rVX}Fv<aqM
zJ0uZfs_2WcT)SL2dyp>!;*Gx8M^sg^jj=x7H#KP?K(Bh?!i5z2Er&dE8<|xO9rA*d
z^uWIkc<?Nyp#FPuRx62gk;oJjND`y>@%25w4W0f_mK-(PpEvNk(ACN50+L7&ha6t9
zg2c5KGxbS%p+0BtUvvn6Y`4{Fre*NbQaYz(A-vH$_USK1YM4XVYngt3a}Ei)18}hr
zdP&E?T!mH8S3xXja+#3I02frD)>eVwEsZmP%J4{tuA3s+tiLNbBGg|X{MO|;S|JM_
zhq&~gNG~{&$`etm5J=YXM0y4qRD<ZR3h+%^NUOb6_!QvPYvwSzNz{+J99t7ocky2}
zf~R(!Wh?(A1HXb*!XHz`^^pL;RR$JiM7{o!$~F*gu;>t@T~95qTp+Ce<UZ1(Nm>b~
z1&AGq(o=;>f(=n{g>N-puG~!PAc#VQwM#4}OvzTQ85BoifgtFA9D$wT!NDrH_-}4^
zBTUM&g%W=Rr5`266lS4AABaRfkYonJi8u(L3^TR9{ngbBw<#M?e^$i|B4cn<;&+#<
z-bT7Hp<jROnuCY|p_WHsSvDBcCjuNpry-&f327}QNiN03E2)iMb!K(fmaI-kULUb)
z>C2ZdJCs%iOKg|R{l0AG_`ZSc@5^vsomxQR?c>GVD;NbmW}PsgC&+b?v^N^gFXU@7
z+cgT^oC$IUrb8o|i^9U}0gkOl$|npRip_cqm->7&B~CatnJSmz_uEOBqNxjp(|sPI
zSflRVh28yk7Kr#-C?`)azUt3QBrf-Ujhk78AiYV5vkX9(W|4=xlm-)3Jl4VfoUS3u
zs!gPlS0m+FEA(`?IEDp#$@TVn!yPv^eK1I4Rg9)6?M<{?)+`{b;Vlyg$i*Q$d;*d@
z>;sEF&|H$kjEIcHC0(86Url-SzT2(e&Y)u%*^h02u-LhE2R-uL+}xNF_EJ>JQkoUn
zG#ZU@u8^S;bn^DywxK9*+AOcj>#cL{<RW)vPE0>AGyY=Wmu!zkhuXz;Oe6b~wX%pX
zk6+1bnIfwTs{$Fsnpu7}$}5qZqOG@1zLQJD29$xX86+C8PQoDxI#c-UF1)1o2(?Rp
z_s<2hs8AO7o$G%w=G@tIP#5NV&Y#@4OiQ3@zSAXVCnp>}=z6jAATl>L-p4@6XmG*T
zPi9$%x(&<ajg#y&8aBPss(B-9N%WXt<P!xz@(>T#0vl4s>a*yA>1YO1ZTj170lWU@
z{UR5t4ndHAo)@&7s<=ii)_yZ~vDV_7X!id4mr4_muF~O(e|jWz6p8@CuZl$;b0qBp
zmpRHS6%zg?niYq_9dLgN!Q#;*jz~HNI5dx(%g86uvG%LWdtm4*qD(VIT%>j={-#^p
zo;lohCwKL#RYXI8(AQKkgf@m=ht!V-mk8~l!&|7=;TCO$ItY-c5updrlsJkTE6d6h
zot&}>MTcV?k4mcf;=CM<+U+Y2m?6Ybw?P|1@TW0YD?yca4r`TOX&c2VW=ZoVJXq3I
zooHEzCGWjZ6{cOfUH(ZFnYvG+ZMBG6B!bnLLoEU6|B<}mbhizb+MN=E_uqXGR1l(D
zeB%pLXcmOjfEaRq4-snvFpQ*u`f+xONVP;$hHfXN;d8|S%cA7=nx^Fbv7sSo91jrm
z?yFb;q*Dc{C2Mg5EzJsVO|nu*>L4nr%q+tULbkYdyB8wOudeX3OuaSB;Db5|^Z7G6
zfNfvF8DMhkDKjvT%uMty#6v~vo3yCnv(`&FoDBd<pB(foBym#w5L`&Ke)m@lp(*@!
zLir)k5Cah;0#HGFLX2!{?DM+pONNj@(9at}XoO$Y%Cs;Z@U^@8G5d>53+_P!CNCSh
zU*Oi0A7J&v>L6!0EYt)kvGufE!*4S}S|m`IaAY7nMJLa!yDFTtYUq9>OxRW>25m<b
zv;&e#tZRSHaTvfFm_mdnr~LEoWry6!g~pSJNFRy$WZntu`t-Dv(~Ek>PP4)}0(cvi
z`!6<sm~J|l?K&hy&YSN<0FUL)B`n^8sh>~(B>Umn$|<9^5Pu(bLnzhFlVbAOg2;5(
zO31KO-oP{t4KcS{{M2lU@r93ta^ny>>k)hw#ur>033__^cxd}LM3O>OzQix)@Bccy
z(VzWhe$#hc;SK~t&cBCks?KCY`638Ta<7<~j}WrSA%Vcaz~2?`%gbv~&Fyh$qL(ym
zaQsm~V^+nIsjiyj`ovFmdrfPkk(p0(m%<qyx1_;uJs*+s#_Cgms;&fu2^1w7e4ogY
z4qe7Lt^>og8lob%oHznt9#cQoY}{`nO{NBQPQ>T~D>2Lo<=G)OcVmalO<fkffp%
z8H$19&8or<H|E;^K_Vn{jp+P`va;vF!tl<I9Y1c4y3m;K{PJ|DR5El$IK_F?&Pvz%
z#(l>Eo12?E2|>_yF+Adtvsa=X_?hq~62KueoG4{rR}NkMtp!@>@QS!!9y!t)@t3_Y
zh~#RJWwwW#YzpGy0znK2=K$G8fc53|=Csz%^|?ETtWxubeXuxf3tE?ZIeQPUEk*`7
zZ-5kbtdtDsP9hjJFx6=kl=KVL+)(qC11~lmbwlu8gXMUIJ_d_S9YS~EfH`LPu&LZl
zn(qU*lFTm1Y2G;g2lew!N;0z68t(5t0@z~%f4V15)0}I*dvV@*q|}Ju)7jPa(7%qR
zB2lq|ny-&$(Jd$_*edroa&NIHhU^M3K#l{_D@R7+7+gG#`lMhw%*!7M5fDhuHz8X3
zQ;P+pl7y-V8ft{qXAgg;Q)^?zZ_Sr`!17eFv&pG8fTCV}E|RrwC}ppIyY!zx&Yl<(
zR)teB+S*J%@)s#~vrc<{#ty!|d{YW)W+HM3f&BCSw-Cc3e!gQ$XlN+(Nd6`TW8)Z|
z#p1|8w6f`V;k`{}Y)Ng!2SF(8$r~sX78iaJ5D<=T-dL#T;!uh6lEh;{K>GT!jR_>M
zjbdAm7muPe4sy1+*(9HROpZokD*;WA-61tQc$DZ|P*U3a)(S#RI!;h1v@weTD~7uh
zOzxE1{8ZN%l-j1gVb87ViDNzk^JLEenVfpqHHcYd!--OOb{u^ZHRa#7V#z$<vG1vs
zgB_%Tff2d%O`3R@1ly<qhXA~nHZW9V=bBr5w*B?^<4IkQDI<5l&J!n2D5FP=<mUIj
zlWFm8i=N{^azm{unmQegKqgU<!Lm>S?~yvKl?MauzX?F?&iPU@aqFbqvioKL%WMmj
zdmQe=m4-s}p}xNULQ%bogF;}gLtg^w+q#WweiGe~oQ(qVQp>UVafP2TE8wqlYNp_%
z8nXwxs9~9KW-nkw0_9;wS{%&r%!AMJ17}!Cm;eXW0;7luw|Vnn+;85+P=t#_XW5yK
z+hu5_B31-^^Z=K<1!mIRQ0XSpO4VMRf3KZtN>@1bb3M!YeWzfZ1*N8@Le9BZ`IUCt
z?BsWUA{o?1$*tDXB7rt;ZdM$!!1ia?w=eEXMkzqf(-iJ3kwA!S?V3`Ff(|ke1cFSY
z`=%as81IB|K~u?8dzFQAT`oAr5P`}@1WIg|_t-ULb&yRBxsOfty|p6w8k#0a#}F43
zb$lAag)_s7cL#7BIj6M&z;YALt0^Hb`rdv%qg7}gNd;4sjmDQI{xvl<>EK%AU_@mq
z-40*W)@M+UdEy~iIP^W_RDBqg)#`E*s2WL{r$h?{2Hvq}&v2&aL|2AofsBmIPd%Zm
zu-6fd#^Ov0qM!X&6H|a3?}G9UN~=kZ%t%dlZy^Vd=yldH1z@+&3>ZivEQga>aWK?#
zrgOi1Bj$BCz`q7ciHO7RGb3y0nq_b`>6*#)2TI}F05g7phcOb)-9B#ehl5XD;ivO1
z&tSjSa?^abE+Ii%c@^5G?UK*R0@UILOrf42Myvrc&&tVJ!DJ%#?8sV*ruj+Yrw>lf
z&c6V+mB{fb947)JC2*aWmzNZN8udG`yIt6?%{`4gIN_Llwrrx+RbeUnuBvp?&m;*z
zQU~PFCu*HjEyFTzYOz>2-ZLMLB00SY&}%<Cd5O>qOr>U_+y0jv)-D=Py>T<3ezgcQ
zS<EDye!5SdOdgg^3WA~B2T#(c_>gxp$2%eHGIfT<+Eola2YzA<0cwpf<H^ARWWQEY
z9g2&+1byx}(}*P!F*cB#6Ku*rrEehyf9iA-TFr$oNoO?+Tr%(m#26*}8vwaxqhm9>
zm4!DHR9Z^E%2PC>uu&_H-x-I{(=UGAgGMipg;d+$l1E%{HUZt{%shJiRWyV9US;wY
z_YoQHVenx>HiCQD9@IA~|Abjh>^m_21wGdZP7Rg9h2^su=63g_1GNIf1Nb`|y`}kk
zwstM~CktRI<0EuOY@3aNafRbe@$V1zT+@F2XrObSOitSmkNqsirhY23%JHxnMI6aZ
ztQ=S#JNL3LqNVR;egB->f${Iv^GRuy=E3Ub1rsL8m0fg{th_uu)!ox`L_;I%)ri{o
ztlD_H8p5XmJ#4dz+S(g__w@<cbt_VbU~F0=Fpi}8UTyuFD5au3o3EASpmw%n2Yu6!
zm0e8j%~ATgWXTe18=F<BY6W10V*B@>LvNzIKqATCA&OI8SX{5(zh8nhj_G~3>FMdJ
zy1HYK5werh(^-Moj)qEYS_o{-veMHoa8c7vZ_RUDjvUS1Cj$@eP9Kr<y0(J5OG?|_
z2vD-SIY$`YQW>w(r`meqR~-v(Sqx*b3<4<koIp;g4;Mc5xFxKqX=&LA%=AsJ_UhHE
z;VCJ8P@UaIZB5D_U7nqF^WCt26@;4|0s^JTcuGwTERNsJM=jYbeFz7iR<g5S2UzwP
z&e@~9ASIFpL2kRx?b{t)UE4zzy8!=%KBF51eO$bB=^HHQDyWQjgaJ1f7loKShp46Y
z)&A9{PvR(^s3q|OoNH?);Tw3>r|OA1DXfWajCI+3jC&;_LEadX_oF#GHB#QfKN&=g
zp`l1N9gN2SRaTV>WH?+&tvYhkYr_Gj6`7ftRi8h9-{@T%`R$uQ{4ZUYMT$yFuXtA~
z+11j?F5XN#H2O<L`uOqho(mJ6&c`>8T?ApA1gEHmGB~um9fGDI+=2YamGsB{4Co3#
zo-2NSeohH@a~CgLCX;2^@*HRETx4Z>%+M=1pa~)iDhM|@=}a0e%+<H3COJPNzDr+H
zTF$lk+TugUj?pjMB*l!v<Tie`<#Y?2;)Wv6gD#_CCi{PQ$HZ(y>AQx4J><tNQ+mrF
zlJ@QjWh><C?f^`+qjynW*fZZf=es-wAx2cZzlL+HOtA9o3%8z8RP@?!`x7U@4hd!j
znnujL)2RP`t%82o-;!qaksf)Ry=?#f{VKSB2ROm=7cX9;#0NZkRxa_R9jY=1j^|tp
z4?h<yRod3J2|Dicw6yrx^U^Ep#>B#HFV9AD9y%ghkd~G8(CMIKzcy}+5x0LOAYf&z
zYE%s_?b1ZQ4!nZ<j;x2x_&)EbE4_52i+1WbJ=f-_!}lLP{EkoCv1`{gZ*TemF?n5M
z<82VqeV}g~Qd0|Z+`t{X?i=|Ww+M2}sr_w5{!mB4qM{b9TepskHFCkurDnrGp||0C
zib+V^Zu;=?<E6Q2``^tut04s_s;mEebhqs4^nuge+FWf|OxJ369zA+=!$FS?i2cAl
z+6Dae9GQ6WAzK{A@J7=*nLY<6q`NU=r*ZkC!zG(&6pKo4<v&{r9B>`X@f!xatfE53
z+FB4YS9RRl&XsIl*Y3JRM?_q~;ZTYyFE77)_paC7yDOd?_Z60xmyb6|t*xu8Mx@bY
zTVlZbk+`3@oX#KGimIxM3=Itln4LtsUq9xl6{18VyVz9GR|rsK*zfS-4&VA5eYDoG
zeRJjb{O5?!o;NXxgfx97{ld!Wi2*%<(3XZjxAhbP5qar_wn9mW{-oZ#efw4)^e{g%
zfDco1!*LNem9hUc>yWlJbgxB=78PSs!4ylNbhdYNTt=rO#%uof_r6SIUY&=&(c{op
zKydIKtiUiF13<!60*Rfs5>L4H2ARs-{HF5O_VAC!@|2Vm$_p#B0|InIfvYISR4n%1
z#y{io61qX_%Eg;)4pmps<FqI2<YgGx&DByDs<__1duMY(O*DPg;?37q94R}FNKJTt
z{sxfkfD852)vs=^5fF<`SWjvSer_hmvuemsai`tVM=Opj-prI37Dgcs^lbvxzU-+}
z*D^C5?l4-Vo7FL)C@B}X<KR|Rb#;7w{vB_ZKkpFn9AbKVXXm2r+qYB2xY&-)&TFI>
zP7154s_L7W`Jjx*rt9U{(7R99e$My1<4b-~`c>J8^c!i9y+KM{4<0;lLJm6kpb3!L
z;Q09U$B)<U+O_MsRh!g#BcoTJJ|%N1@00rbf~7j3h)L_}aze=x#l*cSD_e{<e-y>f
z`FJ85dDWgs&TU6zqhsu(9vyJrgL5S15;6s0CV$=V6OO;%=n026KjjH_&9QFX5qWtQ
zGHVF6I3MT6uq^I4eQs$ntO!Gh0B5qTHzSHB=rpMB%PF}I<O!t;gLTb1`uWvE$bgZV
z#?uVwC@9qaP>wc>iE#kewm<VDV>%XH7?V!Z>8bg2I^r(9zNsm%Wo!N|WMJgd{ZZ_z
ztE<)Kp9Ti5LfO38R^&2fxwi&N=6e1eUOs`ZU8bJZbE^x4UKQ^b=u;38y5N~G#nl)(
zQ9ZA^`s#S>ljRX}@v3?=v=p}9*zkT9or9_#Qid6`3!%JWuy6KIyI`qANOEzJN6n~B
zS)p!nP%hQCbw?xC6m%aGc$E*m+)QmJ@A>8W(aYCdF!oi;yD);ITtX%l8-`|le3&|a
z{ycL0`0@TVE}(lNqhn*R9cgy&-rahs$9MdAVfB2OXyc|SnVBtZ*U}IA3W4<1J4t?D
z&U>B)hs@XGVE6a0FHKhUA+3onkvQBImymEoO-(IAUP~+Q)t-He@K!ko^^FeW4K8?g
zuJ7}Ai<}FvI`TD?q&TPB_&Rs!BM>>-*R%${>)i`Q)*ZA-$;tbp|4dWBa;C3%o&g?a
z@#^6G(DKpV`sK*MG`uic3Y|y@3GdeGKieQ%d3<7GqHANs=0{JS;C2}C5FHY)M!mR(
z#7hJc1^JO8i;awo8oqw*cv;CB854`ENIy7y=kI$w))d-vlOwTj@*J7@72MKHQ_N^J
znBaI$Y)UfG^34B-zsS-7D!zqa6Ws`6cN-5;S$1o2IPG2QQK-4=imu_D(<8@4M;2|}
zyO$N>wsIf#^kNp4A9v#{K_q>eo3-P|afRm~32j(a_X)`LZfK}Loh~{N5ZonMRZ;nj
zL$Rj%i?&Hg6~0?Y&dG76+ZHdeb!+sx)R)@2x~;{gJsN0<&^w>QzL`bql|6l02I!`v
zxA!(GZ1I~nR~@AvU%P(&bzPmBS<c&<8s*<jS<9hOb=*&KfV>zG6h!>06X{)@onbiJ
zc15BUM~1EgZ^`b~o0=XoN2mA>8}J)m&e}H-85MO6SayAGZtki+WcRx3Q$yPeJJly<
zNCgC!)-V#<TFKhyl*n5W@<WjZ6nH&%z+&Gt1$KpDk1E}KC#BgDU?w4G9NBq!Z$5lr
zqO3tiuU)&Q4<?Y${RO9`KCW1`>I$N^9ia=evDq!<<;<YFtF>uJT31X=Og5NnWMo7L
zU%X=Ts&gOM&C7e)$LGYB*!VmgVhK-AS4}9$&gOV1p!X94ogh&piHOc=^U!^nu*GO7
zY3`K(>01jA;o%5daNJw*<Gn(Uxplrmg6}Z;%aIrz${ZfHAx9p1Hf&y0x6ul<@j`3<
z{u|PFt}-3EEGI9|wy<H%8b&a|$;nAsR(aT>C9uAnjwPib?s3y|9#0fl!OD6Kr+s{y
zo;`b}Z)|*hw^n95l!fmr)4t%z0BJ@a*&CXgPVPE*!P3$Xj|}j8Z+2XXzl<DR_Hta)
ziHSIjkWzj!3Ou$YP6;ulI_+^?@WY9J?hr}WZ612Mx_+by0gfDqsPh3Tz<uFg;OMRD
z20)p|&CK{VY}g>S?Y{hvHU>jxBp|&oYfewG^>^*wy|s|*gM@{J#n3=0maSQyEYEA`
zmts7L8;|X*GCD>dR?NY{anmzI@Eonj+?a-#xVTSK`KM3esj16BQ?MvTZl2kRwOaxe
zB&u;{toH^Ar7c!0gvet6Hc^f6d;uxmZC2{r(MazMA&zR*xhrbXxConZXdkvd41nkl
z-xYP$T34>PCC!2Jl(i1__VyQ>vjdL4XpBE#g5OyRRX%V>)lScYSTtR13dT6Ovz-dV
zJfokmyIIvhReHto1}On+beeR_7R`KlYzPJh2J9)i(rYi2(rI<nRVk)3M~=`kv#>-r
zvaqlOJbF}y%lg&ZYY5bPCeI;RdImBb?yZtnt?%)pN5mXnMK3`S6wJ=f{!ID(>1zJ6
zAjxjD{1T)etAI}`ZuM*v5n=Ncvg>PP!;ZV>^zIZs88a8O!{$eDXy_&{NBYlTp*zRQ
z;}z#-+7=!-{pQ-en+4e57?MvycxHg#`|~Q!_2g|(t5Sx9n}<(SGub^B;y!~<#c1Eb
zDBWrT5qSLc>H3$?pSM3c<p6$UfRI=D+qc(XcnB_EzTDWr;58mz(*OeDRC@0#LgZ(m
zlYa%je|-HqgI0ly$WtkIaTO({b2zTiwNb$shfDQAQ1cD<K7xVt0znU)N&P6nbj}wc
z&X<r{hh`uy?lb?gD-+QKzj!k%h=sev#2yK~JA(#|B+u^N+W6I2poXXg5?yvLBy4{&
z=ebS2n4Ci%>K&A~l5MA-?I<e>fBpK1x}qYN$HL^0A@X73c=o`4zxZ{ithDT_SHEe@
zwD<^Y@CK{)aY2Ek`R5px&vkW1IJpdDwM$Ym7{|{-A-xNln3!bhzoLB*5>g?do(!kr
z!Pmyd7nn3*tQccM!z~S~eIj)3B8OapYW9x9Rb%$7mX>gU>|2<l75uvSWun@<aTMn&
z!tu(pt)-<mppORP!A5DwX4;-omjLNDfDqs-^bTi+&p>f_6cohvxw`rsVlExePCJN;
zi~mN72|veee8AkCb#~q_VckykVsByugfWSpLI{(a@bU3o#M4JEA|YGe(h_Vd{*_nr
zvaSl&p!VUzG)?)=!Xv}O2H3mC(74Yba@(;nQO<Q{e0QGRuRyM%B^t-E)*uw?gTI;h
z3OV3uO<%yL92Vx?Ghv7tTUdPBAN<?b*Y}K*lbE4z8uaX#l#~jj+{zI>d!wzb4Lli~
znYj<?b^~9bOt>dQ1D&w@%d4vHnxtzgX=s=L3F1!h0G46|2iF=76rPDsOdRo5-wBc3
z1X%+75P+fqVeGYrzAgFA_kfKqVU&E-!|GWj>2@e$4-bxxUQ10Cz)5fC()gDO@|mXQ
z=1sD4J*Ur}U5j1&8sf{M1>dvL(1rC8D?;9zaal2HIr1TRHrWBBB5q;x+MeF=6nrGz
zynY#FjRS0_9tqMwuo@g3T;b2k9!M5rSR~<GIDvN^toY)^i~1%erI<}LuUnxDGnJn6
zbOfRvLiYP^P*91mX0l!H7Z$+AtG912#gsAPSzOd12Zw#>>FH!-Q2%fglO}&=PW3ch
z9ZpcZ`eeG8!Z}UEtLdqBKqL_A4FGLGbhn^6Rs34FEO`;PB2L%5sg6AETpA||cBf2o
z(LQ=d3{l&$qlf&ItXTW5B*n(wK<0Nyl^2hkV;t;1!WDIRb<kcBk$`*m=-ON8lWaV(
zQvpui;xci9Mj!Pj+ihwO=sFjtX~^TpOEzxYSn=z8meY`dldEea>Kuodb(m9Q3LeR{
z)J*YAG!)8nIIM6c*N&HvESMgvit}zIBZ~MItAcOoC`JtG;OPT`$K&W-<dtZ4=tCG3
zbNJ0Mz%j-xAfWIq2polU4$p;oG7kezdyup7%E}Tw9dh6nj)c7uu!y~|rW4iV8_ow@
zj*sU>TLFEep>E!~wX*hrR}j7+u-FhX3*ZL3-d&ubC1_)iMPn)C?*=UK-@ktgLZ2S(
zsar}RVSVY6Li$}o-eeUNN+Gyf<AB1YJlF<KX55>na9CszN2)tu5re6+SBf~}0PA7-
zrE0(CXx*`Nt_W0_`#iX^`Z<gR#0VJ1$H%*&#n-MF5A~S7rSf8W_1d*dFw4zYsE(|U
zcks?F45$w%1QpLqgulp7tr-r-IKC<=p-<H-dWZ1W4G0r-6#K3-Ug*3V_S;;uJ@OSB
zZZVMbBb;W*uIJ|ASp>g-)r2)u2>@jp>=Zl{t{o@c*63ZsLDewuR|F7kPVdMAQpoq1
z)BGmngF*tfeDrRjysEd*b(Gqj<@1XHkt?w(y<xlGs%%4>xC&ClRmA#b5h|d1-3G59
zY3_g{%~$dr2RK1d{c+%V(2x7x&6_v<Ej>NiSi=>sEqxIgxdqR6*cn3YgW!Y52}L@w
zv|!ILBg$qya)B=bMf@A$N3wEqm(eHi;J3>Nx?M0el`B8aTZtIz-3Jf0g@jhWdPPIJ
zXJeMta%vLyLR6Yp?wc2RA^kLqlLz?ySHON%nHE~lBwg7dnF$IBb!<#k=ah20iPtnV
zHXhN@DUdmt!nBsnOPW8wD2i>c|LG-lb#^>P=hllCd!&yYyL3Wb0>z@LzW$-OGZ0lk
zz-8<~+Y{TKVoCV3JL3G&F=TS}EkC+vcjrBLzz)aX8!MT1#W+1&wvS}-a7wydgD*g^
zO+$|DcC>9F@ZZ*<^MMZ@FfcMQM#aVghb|)UlPC|BA$Oopy-NBL0dNY@|CWCiICwjZ
zx>coG;PfasuY_aDA@>VHENy6L;M%ii737ekn2t0&&`mqffd#_Uw~b2Mfb=GRKDDwQ
z`gZK5H1`3gK?*Lm_l+Bj=@}W(>^l@$N14N}=r8DmX~p5>4$IP|OGn}H%)Z17LB)eG
z;W_BHT^Ot2fakn+f~2VEM5JWaa11CYWDxcu(S7@55XG8whod^_yms^Cpp|r)r|^J;
z+oT}EoAl^qVEMR>MHHZ@KUk~n>Fs$C?ZWYFs<MwC@0|(3lPl?Q8fU}3x`Ug?1yQb-
zqu0N|V#J!{w%jCcw<ATj;DgZFK8WYx_)ToM8`rO2&u#(!|Mpn^*ij%mZ_oi0;+~!!
zMGcJ`IXU~$mETJ|D2F&a)Lp$LDkg>--3&bfdXqHF8LrTxxc)D%@F<9|m>2@++|F<T
zQ!m0?&UJepU55t)eI)$jVoNR?s^sb{EiJ7r$lDvfXoUO2gvsDh(xwK!9PykU<%Qa%
z3@nB=71fCQEs3o5?nKR9j40T;SIrPMy+atT|D<7k4nJ*jditB*!t@#x{bG)zs;aD{
zX{-^@z1j67&;%f%?9-<bVB-=Cbnd~Ck)!CZ&w_)~a6<Z3(h_kHU234P4i6wY;r4VF
z5b>!~%$T)LJ9V$FP+oprmM8GT&ZybC9M1)hAX_6dQ_~I%ICvNU<guWvj?e?n+##v*
zi=_^<CWy~plhx8%54@o4C=AKq?UpOvQDI?+zG>_HLKC}^GhWMp4Gn@Jvxt%L_V$iN
zUU<mY>$h(gArE|I2Q@c0ch1CQJ3oa?2Sf!L*zO$b*BioiBQO;`c#0u895{>`a4CW8
z!s8+cjim~Y`au&I@u}|=de`3H&o31Z-O~^c1^BjdG$`w|H(6H@Si)=X3uvThuRLhA
zjR$99i@`k!zn}(~Vo^?tD=U{na((u>Zi68+I_tNH0jwH4Ghl`P9yYUq98?SsoJ%E8
z7eoz!^QMo_737iLj>+!0XkoE(-8vs+muS75LJ>FJUFLkj+=uB9&BqTPu4H6nsOI5+
z`rEn+?0X+f`Y)rf=Wf85HvnIwC_K<*a%yTx43+OR#O>ucQf|@k{P5ttn?BOq5EG6<
z?Q_$b!L@wLVfNk^+d-aeI^=;f63PpARtV_h=?KLr_RkR!$5w<w{SEp*0sYN<yQ$}Z
zlFWG89#mcFOP7S;!4I~-q62?sfZX?EW9qy4*%5`-mFuJr9il+?s?!XQ3%;wuqj4al
zzWMZt1^Dh8fH0Cu*MoyMS~h221~HxmMt_H}YQWQ{JF{tSz6et)rZqM*W5Savb|KOm
z;SN8Jh|QQFv8{>wK@>@WK$)akKM4-Bw{uC|I^Z$Of#+P6RaSPbefb`4bAk=$o8@gw
zFV!YRoQ2wi6Dc=Pt0=F81U|@xTMPA<$SEnUM0LN7N0wsF!$6-N?KKU28@A%L#)%V5
zSZk2nX;5hF23mt1qcmY>tVImq(#TKwp`mf4dVOMlD@g-fHsI)GM^BFr{&@559RubP
zYvM2<BFP~CIGQ{BuDF;27Q2?5yc@MDJURIe1oq<AR?gx)t=D*L!C^~F{;sGOygPOj
zgF39_<n+Un2jfjSukAVhJ@%(?63<&_7B;1G=QiVP9!lzE&`>7h!u*&{q?d|;L$2Ti
zz@XoeCt`dQeLs--VkV{|IR2Bp(*s$nD`yLhflg&*We@4=bHVJo4$T*n>Ju8=1QC?z
zYtY7QPxCzmX-1`{U6?FASXJom8qZinKt1o;&BFqF_l`z3c{@EVthr&Uz*LRtmdY)#
ziRKEw>FYb&J)XdDhL;v1phjir!BQL*T*3CgdOP#5oYS_C--<An_B8FQP*IbpY2y|`
zqNr(XQ!+_In5c+RcOg+~3{Tm$m^QLZ7@Cw8Q6o{=n^b0uq~)Qk<NaJb$NS9jzVGwz
z>o^ogb>F|=@4Bw@Jip6%rrwWD2DRY!e8Js&_eyWvP(jAmnRC^HvkZ=sB_s^#81dp6
z#A6w#LG5)z+;w)<^Yru??_am&PjR0U-jP2}UP7@en0E?Cl4;-YsZAU8Jrpcm^&1zC
zGB+;;x^;FxnlTu~dDV<bMHkmB&Fp=DQ0;~CFs;5u6ZD<`;jNzRW@7u*KF1sN^@54)
zdQxv56dWuink;^U`rdDywrDI4dT`#hLvx4`$%z0|7?9Mv!LDF7#k9zfES%er4a%~@
zLdWbg^xleEWb4kIid$V<H&TA?4$jTWD#mqaP<g4RwB=Y{-eB5@J72s|U+_naMOOdW
zQP?|k9%$(5CQyCJtfvfHmWbhcGeh`zJag?Y{*W|MsyJoN!;TLnh#yhBFuiU&ZLfYD
zNVn<Jn{_wOeTYlr#Oc$j)fQ@pRQ^_R{h%<>DN1fEf^ha33aaWuF8XF3`-^eL^>yf;
z<(aqef(Du`B;h2SWhOpLB?guCpN{%eQ`4<oF?klYhKi^@@@hvV_`atgYjkMZ8da%H
zlO7tK<HZm;gJ+KQKO8?k8bzbVVM@#qx!Fxv!BNJxvV*WGeIis8iqsC@nFmjww&@r!
z<aY{=LgOT@^Xp&a9Nwdsw}$$2Tkj5QIOm7pan<V3g1dL`+AexADX1pJ`*_6M?C#!B
zSpJF5-zw?rkLfD<c9W=QDA`s|ukXxaTAbkMC{AvlTWw4NNkC<!_DTu2=>7_`c%gCK
zovmi&JFUl$Us-I@ItLlptgEP7F07?u2nh_%K2ra?Xq9cQ2$1Cv6(_>l=Xqo=nPWfF
zu`|T^CVX)T!XfOYq*473MnQ@}$oX*vK0h~m!Wo*PnJ%W&?vRe5#ZFOku$Kxx+YXV~
zG-Yg+>|h_Y!|WYHD=Q_wsgwdLx6su6W#p@FQ`#Y7soJ2Z4r6DEOGwy^ZNi|TDo<zm
z+y|{S9jfybx1^;qgs|+*)j{1{m=%nKNA{CvczNMtWQS~O2o8umm(h+*@Af{u6C+40
z3e62NWyu5@0r|m;`i4CRt7K#Zf}QOA`Ok6RjJ6H#*(lYH&v$EXX|cw|j%2N}Q{SZm
z+PxNrkcVPqY1zK(pZl-G^v;SG8x?H;=<KRk0fQ(?vdu-0W@_CMr8kI|rnj#gKR?ZF
zzQp`~68$^qn{U944p2Gwi#E;=&fI~ypF&I<5V!yP`b8TyIMM-VsItv($&$V{Ha3mP
zaXTJsoRBc`GP9y)kCHPfRGt*fOomUO#gxth9kp;GQB>^*zr6ed<xC8A*qol8e(u-%
z_jQQzoTW^;q|+2x3z(eSEOX?(N-Xz5oD4EIH)s5p8U{|;$&=Qa0q3aYX+4T&n;?$P
z_*Hd82kiPu3Z(3Nw<w}C#>9*i3Umy0C9$;ptUKzgySuw+5IGDg+ppT-;p*BKV?Yeq
zY@&%u$?e;kVr1;QAa&!|ItPgLp3aApDl`N3?%iu)YP#9F;x_D9DmkSls346JjaSXQ
ztE#G+x_7VER@V$KUtdi@s~9D_HD~7LRwe4YTyr?+jv!p_K69pkC>dvnan?Mlad;nx
zUU~9+UG^d~(<c&%(pvF50_N~-3GiZqZDPo72^S09t_A2+>xi%oL%2MeoqYxTQiG~e
zO3w0KzT85@bz-{yH^Ax(k-eikQa)qex7nMgr)zFAJF1ky2F1kv8e!MrUTw}W8LP2N
zOK|ODqX@^R$cPmwOZr4W88t8-{$Ooz`K3!T`ZP|m3ajUn_d`QN17ON@cG<2VBT+2`
zF2|p0@%iU_A3b_>l31ysu3pZ0yi8{GU9)CvM~ZErpI<Q+nfdKKm|3jk<zp%a&E1Kn
zaEBT+6~Z7iEG!TgiJuUI0<N^eIs}5B^U~6!!SkW_#FsGxHznnT;O8tsBg(kTQdWuo
z7<#q{F>0BFa~AQ}TYD036bT=WSr3hERan>-Qm?pUhctJA-LVHyv`ZZ=FM&tdtwM$z
ztZ5Y$dg<tu+FE5c9bhPQ_dQ`C5U8GkP|3UM6Io`?+I8^xsIx9q;jpBWI77x}#>E%b
zb-@k9pyzo1j31(H)tWVB5C_`!Up_#c5o+vX973*dASe8paev~nwv)V@o4oJoz#%%i
zy0rj5HFb4=gwcqer>O?FY!wL13I<>Sv~O~`ie|v7wQDz1F$z-6VbQL&F!y~G6nrkc
zW9ydD>!@Yz`9M@*oc7~qB<z~aiHRP@KjvJhtgPG_a<-sgMTW^*rgADeIsFoR_AI0i
z@>W7h)_0HF+KjRD<wyoADh|7Ev-i|txuhh8!<m_zQ5l=iAEfLkn6>K|r3ULe?++iF
z%It-bn>SS%7X=g5s=KxD$8+airsukeXhYoORYTyvPY^&jv3>Yim3uhhn%de10U<s9
z46eC$%_FD%UGxTN>Xd8<f^)>>g2|rl?mNnwtdFEPgpU>zH}1nhH#9U55YGW+#IeR&
z)|q%~mECW$;jTwww{EpTm+m2@XRjFZRnZbE(j_qVI^WmN1elfASgXA94+*(OBgdll
z9<X~1W&_*BY3Hw0R{G)H$U{}Mu(C=ba-Ad@pIKb~$x&!%cv16nfBdc)*?ac=Uu^{B
zr)*EbC_Z(0|4NZFcLsu}fo_gHNl69;md?%xNv9_So#y5fJ~dItf84lnc|?8jx4NQY
z3pRlr;du&^F%!H9!prCF?7m<bp1|rmcYan@tyd%%p5gWztt<PprF|*ib9JVzXFncI
zPyc6aizkTud7QnSofQwZD=v_a<1lITP;)nSJ@F=3wcY$tk;Kg;_A*4gRJef9Rjd4P
z#8+2UZO7XcKhoUXhCsP3da-L!t$~vCNpf^_9}eQ6fPj`!XU$Aa{TWgZSY6q=WUZ=_
zQWOfks7*mQyVpPKz_*J4wTvPsDmT|02gr^wMK7fr?lUo>nF@-tRl%nt85w2v*-x4E
zE6@O)J0TCQ!|%b={{|U({o<xg%v#(M{)3jerDZubniKRro0yr&OCVD_A)zi+RSn9?
z$szv!N+}h^_>Re)xixiAHL-QZl>2a~Q=j-ZPZv%Pr0n?kcr8vAGPBQxnI<X^AqNvu
zwrmJuikkPLq>Jx*=3l!qz~GnT<KA@4`;~3%N-c*O(azLZA4}gQ{1Mm5hK-(DIz!|P
zGxYHPDdTFY*NDQ@f-_;GI5%~5a~qW_r=B~19szH6_+YODW=3znRsY?g%uG`@)wPk{
z?`Hp`clv0Z+3><Qa<P@xNV*o-%df6W=lYr`lNcNvKk3x#(ncqTY6pev3cS*S=3}iR
z{{!G3E25dhq61Iw77T(i`&~DtPEI_7bZiTT<HrvJMfw&5t5Z@6YL5Uwr$o^N-~(g8
zD;)xl3Ayoq^9vfUCScue(zBjD7wB$>V!)rWeA?b2meM!}9AmY*u}<EeKaGhYEX-`~
z88hK*)WGa7I45z5i51w^cXAYFSnP3T^S#B0sDHrCdl9zul3TZiy?=e{qF+FOe1#^@
z*b0Rh)G5aXEHqXz3I}0>zqhhN@)BWd_`f0e3fYoB3*O#duunQAX2O7F2F$qyUK?r#
zJZ}8O0Z(YTgV(W7wX_!FnNIoa>v-2l%xJvwpPx>&N0Nyrp^SL`iGs8i*s|YlwX|@}
zT_zLJMHtQE(_yXWIKail1>aWpjyon)#M9g>-dRgiv*!EYrgD?V8mlf+D4vGOf1_(#
zr<KE(0-#DA<X+{I(YEQm=ZS6G?Jn3tvVvailA}RRfsi6<@5y}u)i6R>Ww9s8b%vyt
zQzK)gl6_I8AdEz<yV&(x;NksBWc^gyX}FmR4`!bkeAjvZyw(6z4Q;D|*@*`a>ajDB
z!?2!uK}d-kZf<8uFe`p~Io9^aW5<rob_(psPksk<G8i#p07z2pEaX`U%Z{p10+zpQ
zyf{YKuezt9phR<OnIGwoP_xL~#?euSf!9sC`P0^K3Qq1FQ>|arKX&y*AJ?Hnho+2Q
zsmCad)wFKYvFFD@Zf@s1yu@Z?KRnNV@-u4;&!d2vCiRW^U%9v+ay~FAt>S}gEx7f+
zQ74(|@txCVAOVo`pvKz084&qhHUBO(%lfUOUKYC;!f0Lj@*4+THYFkt8)65esv0nG
zU^MN35m3}lh^5L(O2yWhPlZQu<Lfos^lT)zPcK!Z&09v^=o;uVgRakhMX#TjgLQn;
zHnLoH>$*(@Vj1~msW4<o`-Z?{clP#frKEIy19>h4dTPOMr%#bIG3@?&U0p1~lWQcW
z2)W=$TH2+NU$*lE1dZvna}!}Io!$#zXe&ZIeFzE2d@Z<0w-N^=P4DT@PWt|P8Hs(y
zyM-e^+1nozm?dEvo5rT+FYYXTZZdEs!Nq&@t))RhZpQA9CXk&n<6R`6&VnXBL)hkb
z@A;D_*w6m;eaa8mRScF=u7~kJR)9M$!F+jF|JfMvT8;k)oK#f|Z6GtlX~9VR6-8|*
z5!N%*jZz}7lD2l0y`iVc%!14AwX<<;B~PQ6U{K9Y_6;7-X+q7wG+*uDBW9}G;!RY5
zh<aR~JB$Xr_Io5>@jU6KDCT9xLenD??1>m2K^-r9+K1@<_Ho$!`S+$wzj5w^zsn66
ziGC$)1pMj|e~blIwziLSx4-=Y`hs^bT|F@@G;|Wy&UI6bU5|4VU551j<+gsZ?&MjD
z9ZFJXsHQJ`@dS?xxgSTs6{X4IQG+ghK?FzdV(kX;yEPhDyQv=L>=7nZIv~W9@C2ZZ
zmn2-@)9(c?${c2Bj3PRw!6Azmt4P$Xa!|>ZA<Z$WU|+RSPi+3t8vz<0q4Z|nTX_BE
zQKUf2UhaHYnyXXpD;i)EQsOrR(LHEoV^fNEG-9gZ%wni3lg3TA0|HmBWRtx+|9Ac1
zhQVVED_#kgM|M~Dtk^>^2>yue1kowvmCS>0-q1}!9*WOmppMrHNJ9vz9D~Se=LJ5=
z)?3I93={Nn%$lzJ(RR7OJ5=HWf)&3x#`vh=%-ypis}ab`szw#nK&oJXfrD_eGyTSQ
zF84jaFr&h`Apk_Xg@$8~E5U!f`&vCZydRd>;}IT0o;NUz9g%c(&gT!#n+=)lxxSh~
zhe`YPZ6`c$zlS?QfP0xt4wFFu9ayuyy&t8VdU)MyRjg8)gfIIC1}eZDXXWOWg3Vw?
zT6MRNI83H=T>pFk;w4t&$@AuEA@V8f>RP5WD@q8`rt57!a<9=4WC4wZ>+DuH_w~v1
zEwT3h*9jd>9AhAl6P57lrs6)JPJI!N>_dlB#9wyM^HlneY0Gumw1HrYu(k%vZOT<h
z4yD+~Pp)nA!L!+4cshUr#o~=Ti*XpOf>v?G5_!};)hI;UA<P0e)zs0Uh=P^)XIz?y
z5(f?(c)4fU*E>@$TyWTtkZ=<F|C-0w`#>gdh-sLO2LJr|^Rf<OBcmp0$M1flvO%Bf
zjB6N?#le^M9c^wdidL_7K9EbAlj24gO>VQk2?c{IL~<AhG|f5W2$hvJH8oQiix*M;
z%_^~Wam7@O$F2_Qfa|G+j32r!&dBR0JsV19ky-H6R8WaRw;P_U{u-R^{Qhld;y`VU
z2Xez>m7?dLS~8IKcx869wzf93EhF)SS|NLqY}kG5)Ma;$$NhjaORlJNg&ld30-Z3J
z1ug|R9L~<(My|@r%8EnHD$>P&)S7a_zjmyZSajj5Lk=)8HI2630`It(2n&PVq?<4z
zj!|U>hK7B>?Ly?FyD%v=b>!&1!t;w;u>};gf7jLDKZe*J<~{q2`h@_mGWm!RMlIp)
zI`sVQtI!?V<A(wvoAqp<*t^0aBO`t1&BI?E6)@ck>7EBO!P}$qjWx6?m$xFrHDhE@
zj#FU#7Tli$l(V>);IlM}iT&!g;|RB*JV#j?8zsK#X~4{g6qilT9{cU08F?X2qHvUV
z_{J*PPr6yGB#o1+tgbGjvY(Z*Z(j>w@Z8!q4eliCrE%=sZ6Dr2Og(x?7lN=%Gvaq4
zxT6JH=0jLZP%^q4gu9sQ_!|{O*s>;SHzwoL{%Go+K+Q36;X+-K#6Pr>Uls9;$1WZv
zjZ+dKJ$67H9pTRQLk1hh&i$7KRJvDFtM3O_+>J63wsFP%zx?B^%e<EAJ)c<wJ*BH&
OGHtSl>v5@{?7sl98O>+_
literal 28979
zcmd43cRbc@|3Cgl8fX(0X;2{~l@&^nQL>5bB3s#8yQD!W$*73zl)Y1klrkbKmA$jK
z@AK%oyYBBFpU)rP$K&_M?|R(#bs4Age4p>*b-Z5Bb(}X;l#ecF*uX$h)N%#+BWe^y
z>q}8ItxFf<C;PrMwc$UD9Ap$Umf|1JrKhjs`(^g>Cmkqir7`(G%?s%m3;a;jQC7!M
z-OkL>#rXUg%Es8y-rCO5+QMYB^O^Gw7IwD60zv}9yEdP7bhH;26#Vb+1?<k73;I$*
zM<{ADrEuhshHGeFvzO<cv4w?&89s@~`ePC)8%ly-w{wYADtua_`bnK}%z2I2vKwut
zyjK-JDXKMu_FBKtiQ4*6p`tu>`=fg%#~&F7e>3<lJ)7&D7xKC<tAAf|<JfsOeSzDZ
z=h!Pde|?^5Zf_Y8z1?Y?qMN8iLs5kmO<MKzcn?pER#W&sl!;~)ep<Hb|I=?WnVo4C
z(%nYP^2Qsi8q%yD-sX^q?&>-#IsAT`b1B#M?LAjm4L(O7e|-Pmy~qvwEW6&Wl-9^P
z-_E7*=+UDMlJ2Pw1PzlMIzPo~#%sCH^k+CtG`oDa=MWPU6L%kdq4;1I&&^_!(G9r%
z;Udelp3&}VhNVkC+hoj5OdYt!=(~RX`Z&$_2h(G{F-iJ49SsYOb_KB()zR+L-7)(t
zs<_<T-FbzCR5xwflx`~L&-S!ZYUUW9R)Wf-J!ke>*6eiZt5?l&?pI4ODYkuaJM!oq
z-rc)D*E`|Wf}Pvm2Q2J;<MgQ5GEH~n>85SF)+r^eFEq$=FRQK9yvZt<|5L2HyL<Te
z?=L@#Z`QsxKcTST6RelxoQS*Na@aH`Cg#J34@yUGv#Z9cJs<WK!b)5X6?am+aN&Zu
z+i=tYNy&wgsTCr0uKakj(?4$t?lr4mOS7s|&T$#kQc<~Ea%-bG;|+3kT3Qbp(rv6H
z1UGL!xX<dVhP8E{CY_O)S@a%ZVVlO76A=%%f_OFksHGIeUiDEbS}V$cR^j;Z+DF3X
zd!2eTBqv)h=@xkBZSopbb?UB?eZ1c$kn4F=)ccPgk3Uy>m@ZLQS$Tqohex|#V(V6!
z@bK_-&I}(KsxUE5p1v?NlwCos^HbPZlS5^!bAQ7I3FicH=e`#%EpDA5ow|u%t)kn`
z{VL&+lx!ZCINx^tO9O5myKDf1FCGzd>eMNXJa?y@am(7c?x7!V1G#*BeDpKz4_R1P
zIG096N0(Pt4#s)q<m5EkPMmLKrKqNIzZzj*TU*=g!Tg!~_wS#s%eVV36(neA5a>9V
zH&Kpzzj42{%AHM;TneSPIn-rjE=He-FlgOJN7?he(oM{E{r&9K>(`7q_w~rCw*CAW
zXs7f=>=bptp~HWC@Vi~3va+(4r)O?%UY=&sDKW`~*^!=auVX^?Sw%nK*W=y3T`t0F
z{89egxSm2F)-c?A+TzxReTrww?+0XLWXRpwl=%6xN}6qR_T3}~%2%)S3L~pyU%gP6
zq(}NorL5;cJ5^=Qe0UJ_`0?Z4`5d<EgY63DwD|SXD{5O}PH@V~$teb|6FBwa`t_xA
z^D~1C%a(nKQn*)><zU{J(pec%H&Q8`j6eUzzi7_Tpik7P=RWErC_MaAzY#^vOOr|&
zt25M0&=H!NoWug&TDhIgEKS$xM-e@TwD+lR-@e72$~vD^96l1F&MrWSGth3PsH`w)
zAA_>8GG4>{T*do4RqNZEsy>BF#vi`6f@kN>_XC$$)mj`B1KgJ_Lc_cIcxsky{_*Ur
zL`TFXit@d;g`&le*I)BsWTPnN<18iKs!x7#gf{$${&e%!t=hSX@5%dZ8g_2qUSh9V
zSC?UD7<Rx-EnX{;cPK^BbTL|vXb|p^W!LAY4X@7J>_5Eb(^XQbp>lF^SEXi$4oi+y
z2=~;dMB3cQLa{Y_O*Cgo@o!{h-Ep({8Qqc<@_}X*!B26|o4sfH8JL)ATHjo9>Zn-X
z_&XpVV0d`A>Y<@`EGe(s5(A$vt(1CUT6$X{u;}eu78aK0wei{?Oj6#Pq||@iBOt%%
zAVszA)k-=Qd-v|DE0-^K-PjTSc?qZI=J?vUM%$d@dG2FXRnL|BZyTmtKSHmH@?+ss
z%pNrx===6M=<eMoFJA2Qnj5RzAZ+H}EM{IAUfY{okW_q=)oP^UqqyhfOB4sc0_kb)
zqg}>~S4esOD!Ikj@pIvVImP&mR6;#>eXK^TTKK0<_hHsL(>uLohNtdGCUpk(^;w@e
zbB0G)Sop-R$By%}QyZn6-&~|8rTHmBI(Z$xZdb)4VciVduuZP5G)}!=j|-bssHmw0
zmOm6s#(m>A%s*vX;-7L$=<MgsdWNW>+RMw<?)99s{}?QEIPiW@(DQ_ZU3nv)B+`B4
z&;=6=yr<46RB^a8Jw^NC6BbsVpP!GOXP6f@!+h{nug|cHa2{xi9ca!`3dA<9Nwd<H
zm6cTpw5&@?#O?E%{BbQX3VT$!^bV(XU6Q_ymvuv`0*APxO0geH`>$V-3v+`$(=&qw
zK>-2A646I)tZ}Bvv;Ou<@%VB6>q@!Ce-?vz>$Yu9zkkYV#%U<HG@ZLPkoju`zs}2p
z@!H8+y1F4;vD(Q7=~U+VcEvP{YB|zTxVk?-J?!2VaFp~F@}^T#QbNQWqN9|AeI90I
zXE)rkO3`UM%W1F;W$UZgg5^oG{<i0Mn7C$RhF!*-b6@?_^X=spxX8e$a7ho|En5zq
zJ9p0S_HC`oa7i1lcg4lsxFI3?Z9+EeJ^Q3Qh_BILU{bKDsp)=Lm`z`U?c&ORo)}#|
z=)r^Ne6N}E>gsCS?~!+;;!q%@n&VWSUfi(X`h|J8n;fg4K@CQpwM3Fhu^(wpE#5OW
zPF<f5>1Ei~UZiK7ci?B?)%Z+`I4X)Oh(}!x4LHke`N&3LUsUS@R#fqlO|F}@C+|fa
z-Q;#edVVs}d9a1QyUM0D!h2dFOu|L$_s?RR(%LvprRwP8)f??{pLz`DX=BQL=1a~U
zj<D+fUQi$-BcqXQAc>~Yg(WGNV$4xk7a!St-+5Jxav14zRn^s4gjmmw*D}ig@5<ca
z!_uKwh}NV%HQe4lFc2QIVZQ9`+h5U@bAom>toL(IRh}WgJ%9ey#fuluk3FmlEn-5?
z`L}P72K4{-4ckyLtKnHJ&=viX<>z?%(0(Xe`mx?RZFH5k{{BSeFmV=VvAQD^`jwpP
zu`#}Snzz+<J*kO$8FAPdXa+%F6Hb~Yuyb;9${sx$k6Rt59&?+Pj_%f&&4vtLQr{<2
z&jy6Mb*wKgDe1vra_ixfkkCVWP48C>;<cC<XkNiF&qc1?ek(A>blbfXdyNXIG`sHx
zN)L9W^f%_Zz4l`jOv1~>de2RhK13Sf<;#m2g@6qn6ciMEP127Scj|eM>R7jxm0vds
z6+;lgmMvQ_Y#t<N#dCERU14+||7Nb4Zhg|t)%AtU6$ZcS*LT+6Tv+nwr8jW;%%<>a
zCnfM|#&xBXZmNDJd((PL|0`BQPu10{SCf9~>E%_Ms3(Hn@ohXp(nBjbZ#;yQSaET2
zS!wA}Mpl6u44Bd34}yGBQU;_z;*;~^=o#0cuU3X90#5uvWN0K0cAb32z{1Pi_Pp`O
z`h%oXVsnnwo8D0f1p0{g_h<4N`6%MvdSQumY3}6Do8&fgS)fNTtXT22r#7Bkh62GV
zW*<YIdkkciIx8PJav8|a%wsAN51yXtj2LV27_=z#r6Jvtyf6H=+$80N=hQI0Q<?TV
zQaJVuGvoc$pTZ<+^E_M>0&m{DneE(9c2>?WY)GS-*O<a6)MwZ&dKj{gWey*<X+#gI
z8!T8zdiioUi=e?NZ2E`ACSN9h|8i<Aq)|}&p6|uN$(f9{wS43LW4(3Bm6b|)0{=9+
z>;C>Vc)Vt;8n@o#TFapL`1q=-swB6hVw04)sq=+7eZxwdg@lAGTXM6wQ2B2MX0x4o
zdBwy|_I^#Q);9EdzDd&k1?pgW^4FaaPkoU~<mTCb$iZF*poorDkFmr%O8*XEtO3B0
zLm%qWS8qJyfd1w^H>QmjZjbYtnVRAOy8G@q@^PQlPR+Rc?&E#2Sf#V=<w2zQ9DlyM
zk!#p=+&+`a1lX3G9eN8Kd>3yL@Hff5ukZrhnG~J94jspG##2-GUb%9mth!nqt%QM*
zvAXQuj*%QktmqR`llb*mf1~z3%bJ%nlG9z0&KVcBZQVNhQRMsC!Ir!YyN!Iy@AEh`
zUkNjG@8<p(vQO!?*~cohxm5GYO?eCRGa=$maa`Sacn!wm=m)`qv$(x@o;z)WgMxyB
z$)0%w@#iwB;DCUqx@ndc15H_V2W-)u8uGnzXlZHB_xCpEWc>W~i)HQF$lzeM;?h#S
zUAuM_D1w@;5a8uiTzO!~7of+X-veEK^w%9}OYFwLC~X5bSpak`OjMwDA+OorrRN<S
za?BRyrpKyt-9}P9r_O^qc#(C&w2Z=H|1;je24aEQ37&KuYCV*wpQDx0tR<>i@gtnZ
z`g?v}Lx!EtN7MPqUn?bN`d&|Cr_??@yd=l##*G_(H*S1Mw`sIqc}@UB2{2F%`=p-}
zZD)7yY`wJIbd!&?uzACd2+#AiUQRtVU$PuJQ?Wb>YPd>3`6Rwg8`rIKu2p!B3m~&B
z@Vw!Ph55PK0-uG)Jd%c@O-!w0ZEbBkckMb>DZP+uS@R;mYT;|5UVC@9#l+$hZ?7)L
zSF!c=^%r-sfV*uGBY$x0@&4li0s@#Y^(+%H!0~l;BHB60$;r67qz1cxd4Y0Rxkn0x
zyZQn+UO&g#0!`8H=FQh0bJO3GQY$Uf3`RnMLyw2-Et3xeP3*!09%18Y*Pmg&zSW6w
zGG{Pf-O@4+GjInhD{EF}CNSJ#;G+Be4!l4M(<7Blm7CcCJH5TVwfjAqGS4ZVIFT`T
zn_Vo>@F8e`$_u|5CL<l4V89leI@F#7>N3`_z&kKX?$!oS27L+VJ{7DVw*qFfw!3@w
z1`%7&+Wh9`W-@7fih=1FrzR#Gj19c!IVAp943DW<GT^GMtvxj})7UbQsFfJ9$#dw?
z*x+}=k<QA8T+7*ZALNj5Zd+~y7TGoYiW>u~azex<tohq(b8h`|@o4aarYwgT-E?b<
z#!X`9qqt&EL`eDh`{zEDdLW=5XZ-eRtW9HjU=${gzKi}L>~F7iH)S!Rs(=3~<<{37
zXw$8%XT9u`|IUr?rEa3Q$9RI8x_Wzi`*TeF^_*{i&>G!b$3I_vzw^zdB@StB9v)U@
zf!vGZG1?wx<>V0f6&o8{GF|aumo_?Riuf@hA*;^zc5cumy4to5dqUks_unkux^m^p
zlZjgGZEbAbuYvSMyas-}y>_d{NX;PWXg2fa5z~&rJ!E!d@;tO#^7!t<&SfLQJDYbM
z^^tld&re5(S)5Cd5db}T)c@OH{{P+IKmd3wV)qsJD#6#6I=O3IqHbDwx*#8)l3?!8
zMN9@KC<=V&t0$zUqHb@Jw5&Udu3J=8)Yje2bK=Aalw)^W**$tM9<oi!uu9-07*)@)
zk2@M%#`{iUj9L_3UWRQ`9TVY=K0S_puIyYoHqfm1`}c1&2-Rc9{7E-kw@yptA)co7
z^C^ZLUmvUgbe3W-G~tGq#+URPL&Y343cS5Iq&#)d)zwW*o|%}Kq}nv@-*@)&!N63b
zH?-)<@(KzHV1;E9R(LM^)6fSd4WC0w0E+>32uyYA)q<wcP!z$n&$?a(&+T{XR-UCa
z#1vqS`^_tH0a_YSbgfe(o%@{objWo+R}A`x@1sxDrR3-5D=H{tSsDUBq?<36l$2y)
zV+*!(LTNBZZ{0|}e)Xy*URw|q*43Z!J-yL9Mdv(+U7Bukz4y1wbHDC~huh9@h@F4u
z`Vkj+yAVw@clF7WCwsm`UJJZ;Zw&0T(QJXM?E7~^pZO{CB!}Ca(gM=b1@YtE0|Nps
zE)$-yVEKgs%vtLTKl~W1GxVt?Gjz|H^7?)|v_Ig`G}F=*dX3}#jo`gWP(fH7v&iz4
z{xG8|L&YNJW+ocb@7=qHi7seoXGgzi(F*`PvO=8^K1twFjSgNdEiD)h>S>m>qpg=#
z#`!GFhe-S6a}nqPs=3XM!n=J>L)S`R#;Y?*jpOeFws9+XO^>R&x@MwHZc~WU&q;4~
z!#uvFn|gMOf^N2B?A7HPqcIH>197Jvi=F7v;X(u6(idXdyhk&2T1Xq$?*KgIR(cS;
zixCU)K(#HL2E*cD%m1#qu`TljxFAT*W+pMCMXkRftv1KS?)mfQkJt$Cy^05r(SXjw
zt<dn={0=(j^J|;j<(I5rf0_7e2_8~vbef~Et?jgCvVnf!XxWYkLV-i}obmhl^XKU2
z!z=$j(IXm12Zw6X@)Qic#)A48*>RiFAEBF3qKmf)`yPM1zZUnQHcDY_e|^efKJBFY
zT>Xt1CoohqZfwD$jIIZ)FT~a;FD{lP$h70*<CvU*Y*eLfvp7+iwqM1ge7DG%Q6GZ%
zVsP09v}YdM<y$=Sz|YxETXPhC_ntW}m^T`@$)iu`d33a5cIttOmOPK+%W*R)I{!t5
zxCoxTdo?f*>d1x@Hvc3a=sn%N<9Mj(mv?A_)b4lk6gxyH1FtbnoBEfBAQh#L187+S
zj{SXo+Ey^TQ9zOo%L(Xn-+dNrOdvT>Y$jZkyPI2geTr#TgUeuxjmI`<8r1DK@>Cx{
ziGk}6dZLS&r0m?W<0$?rkH|fk?p6Q3a_9lKSD1%?-nRLtP}Q-X+S-JZ!uxESbjaKK
zdz+jneSQ5j$8NQ}+2Icjc^(<(&!5+bQRaZ+^zFz5c^B{HsoVs$=Zb)`fn@@C_eQ}1
zRB%{W;`{gVHcgqT7zgELWhspd*ietzfQ_idsD!uIR%)L<9VPDA^#mLO&^fqk9ey=c
zC{LYAF}d<bPQj9I&qjS8l-e`A?j;cN#Y>ltJ3<Kd^)%z6ZbQogpbM<E-1mh8e>xlx
zpjW7uZv92%yJsX}GpMigPEKi!3*9jhA^*Oxfq{Yc?7H3f1NJ)s)a08boWWh>qUa0X
z6P`-#6vx_0FU+bN8ygc~3fK;GXf}cC@cp&>JY%Na_k05Sw;v0B^6JcoSF85o7j(ko
z>xNKWF<?fIyH?<7rry8jZ!Ii5Xlg~-w&b2%DLwlVknwbiPO4=sZ_b=nyylM|Kk)o2
zpuDxuWEuOLbLs|y|0lqEs|H&`rDpvyUzAV{7Tjs5_053^h!KBLsdkgJ*R&&1M<74f
z;-xlY*b<omf@(TPH8vrErqBbiKmG39(bmz?3Ae)g$z);6qHluU3&~Flx~tlaHQeX<
z7kL4pZybongs}XcKgTkGDg7~2O#N5s9qoq#`dkXCs;a{yBh`5Gd+l2fa&O;Gq{hZ{
z8}&<zmvbpFuG^&&DR&EkCxGkS$GM8QJ$jb{C{ZB#wauppy#~6gNw?9{dknQSM8qzs
zzFucJyQm6!Ul8~^|Ez-Azi?59b~A>DyBw`igp}7YM+beVyMbKN($YE3_!l&q_s;&w
z$#zK&US3`{orVhk7i8R{x(6@)R%i{y%>zaZ#;t}ywQJqYIPK&lG@<qn;=T(8W+j$s
z4Qcuo>qC4vN9-_3LiSn&W^F|ov)>P(GJO`jB~gac0PoCe)@+5mhTFokd$)?A=kF`v
zvZqbh0M(ST_kPd@^VRbJ$h6V={O`!P!zmqZ)iD%--X9aN?Px>#i7@f!sETc5NQti5
zW=FFUG!D1E3cqD$WsPahap?t#XJBBcdTIRj7}FL7d>IH14v1`Im1F<oA{5FfJe^V?
zk@ZGL9ns+4I#>KNV{9cl6q}lxA%F1XaAG!L3o^BWnUc{Aq}JE~J`{<uAE$6{`$B9o
z>gg;ne0o1xE-K_hW1}9%dG+<x+xDI*XT}ac8N{dk^*)cf#<L^K$ovh|NYvd&+G7ud
zn2yQ_K?OpxAfmr##4=HPm?;3>{;xlB`k?S(Y|WhT>Tk+YhiIH?W0=nmfuHEcmKDK5
z;w}TBi&t(+K*1Bt03xdm7-`eFOFvr!i~l;0zX{hFpTvh|*J_O&3C?MWndiPRH=zmW
zq43Z)XT^rS@?qjm(<i*r&VKnBV#t!Bvx{NnN^MuyNzFUI$H%`0ax2YOXX3&>6=83`
zHm}qIfT($)x(y1&`?qfogD<Gs+rI?MKl7iadV29^U=wI|U+3p$f`Ws0V!<H@@QI4n
z|EJm2i~9yLv$H3(wi=x=H@AY)rl_Q(-d*+l_zKjF?}0t%8C~C9UrnwPXnIw)+ek;8
zUPclWn=ycpkbfs3nL5B{+zQiuuPYtF?dL$(HVBzqF1f>L_iu}zN6$=8c1!@%lXM?d
zL7|L}R-YiR`tN{v5)l#M4Sa&DAzK{E(E~2*=lzEH^h~&*$L4L^ccJUV9=)@PS1&yQ
z_1}<cv8}sG*B3t3J|PTsiV1rJasbPQ4RL78@g5Tc=UrSLa<TAf9mD+n&tu9QsEYD(
zHSF_!KJ%VV{SCFYhS-zC4U*&;K8y(aCM6{)%FBbv3#kIUQNxb{9v%$$99C!)_r<sN
z)^Y$?C}0RDl4RUr;8f$5*k<^Y_^;o7mE1iV>~q*4Jzo>~6hLMpnx$-)lZ~=&8_f`=
zuh5s}cp?Vh0CP&PvB0Mwe|A_2s-be==``Kr&g3=8+$~s`*8_T1BR30}fsDw&C?eq!
zzV~;mQIl|ii2ZeKC8rkJo^`*Z$G8Rn@ZSfRQ+3IP$$|y5Y6LXsdthTg>)f*TFkYo_
z6)*<0G-n94P|*$3D~oXTg+dr>P#OM6Ta6%9(BRzr)9Pvf%l`gZhOePV|2~wfjD*WT
z4a@}Ml%e%&?@$i?Jn`M<@5-NE{L+1_2hLvv!P%sh<K=p)p5KIEnh06%zd-0d@u3Lu
z)o{U;z28wnYXoA@;<Mm|#!M6`C>K3-$q_J#{z8)`+Lly{d(b%ddW>5Go^`*wu{H#z
z4}OwrQN8`|wF>(lzRIKudup#;%K=ggfzXW;!TPM;-`#?R-YtasqL_B?p__zNX@Px}
zKD0xDFiAkKG5qN7MK1HjG}Z(|NE(0l?nrHte&c^Y2qq->sKw;aj|QwJRDhid(B~ev
zxcBlCm0Nml^j?2G(cj+P#Vchh$kDc%-jUp{0NbJhV`<kqOeXvrmH(J+4IUv<CQltx
zR^I=&OemdvPq($r_{|fF&5d|Ml<d&}Mqi?2D+%UBJ$bU3Xl!_IH4uLi;5{@7{ZE5A
zV^us5f<3#e0HqAYVla2)cE_i%c)Xe85mNd@r|hVBWGZ1cJ=z^2WO|kSj?;Vc5QGKG
z>gZd_c+B-6QpQ)evQing&2hl$#IYjcIs`a^#4X`V1`2$f|G(y}#%r_VuqnFX*nrHR
z+@g*8R+;ptFTA(oL{HvCOEoAAw6tXkYNt+x63s{buzVoI{<vEkMKHI$0;AAkR0$qB
zyi$^@yJ}+@MyF+CdSXAji;K%XyB6X%X!nbNutPY1HG>Ln^7GH7pRGP29g77U9v!X0
zVng8Z!RX)<7uUVEf};wutt;Qs%@p-d7T_4uHU;+|S6IQNk|5v3>7<yHfBN(lOw`6|
zb#zpeeBkWtY~QC5z=fxWuR+p*j=KGL$ms-Sti{nYrx)A9o9=1MNGAFrx^jI3;BxZE
z$0Co}ip;R(bYGpmL8QmbOi^+#u-CLTH10Q_TX(bAc4k<nMqAjnNfe~A(Hrixkgf=N
zNF*sp@aVH}p0XMSvwH;^{X=;O@hG{?5fAlBA&}5%>}A;_M_xc28J?JU=xDM1B^;gf
z5ojF_9~tKF0hwTs-ujB%MeB9{3d-Hp_0;c2S&(XBJOr~9bzoVt`}fbhx(RHkw|M)`
zo$^?2!j5egu&v2H0+Oyj%TH=9Ab_feFG>(}d3kxDWA}J3qz*!ANgI8)YCl1T+fRfi
z2<A`9ELpip!_hI7ScOM9y=wN4|7RRS{^x_k1NfndepLfd7nz19XO@+-r74)7?se~a
zdabRuH=ZOAFm%9-RRP2Hx(%ORd$=Cs_~_d0$3DYx3<?Q}gTz1p*TRI)g1H1<>3<o=
zx4we6sW1b1Vs2Jp9Y(PLd;$W>=xk)D5i#tUhXhz+4PIcMLq|pYCesuh=1rTDhKB5a
z{X)$%J7LqFAwnoI`|j!E2@797UO|R&{twe@%Qrf|ev5bU=Vx~LQ_;=ikfEF)SGc*k
zDP)hHxx0p25vt%r;<L6Mw4LT}1y*Ua8*VEjw06zfwNBUqVWQ_`pmCB*GH`2Wxi^Ma
z*?h-;7oD(i(V~xd>o6+P(;=hgO?5nW?|%O97!#_UI*eg~`S$HAH5&u6JqK0)->GT@
zxLAYIC-4O=Au!dUQyI5xx8Z|}q}6{EUPH0M{3e_ccFx0V*RG*lYM&gU3*y()g1aC7
zcaAtwUY?$*&i%S`)4j=x$Bx~vwbTTe*LbeTvbGsFsBm*6<00*Tpw|9x+_)?z;q%0w
zrev4=4=EQmM&IvV6Jg=5KbH}S#b@^SH7*NK2tvDKx8Sdgd_igmbB0XHYml2s21Z52
zF@>t`>X>~%b$`1%%@EXzT8Y8jNNDr#Yin!2wq%|&g-G<@t_}Uh5(+}nILI9lm;QTm
zzP&n=1|{NDKiZ2bI{xv0@8<=&C6=V0Lm>y!fKU~OT!!Un*Ov_kTwngZCnw(AS}!bz
z$cg2*SI>zj3<Orz+^i2#{OjvVDXag$pFTa(>%PMK7(<wV4@4FwcpZFR<G4R8Jg}2;
zh2x60u@i<6iIc$gK34OG9|m@sQ~HQ(U;vQu@PeodQx)%(HWO}3(D>RN%E5n$+8=C4
z5Fg8`RZoW7KlF|_+GR9Wnta9|-F_=ic|yKF=7`1er8VNBl)>MRpwm!0sgeg50-}IG
zNH7NjAJ#xMI<wWQKbMDx&LbG%-j`B33P+6?W`T~wZ3={8e|RXUZDjQ1hVefumf;)W
z-ory&b374&0yoy~xc?9OgmlpGcU&^H5{e171joO=0Y>Ov^H1!^2EF_D0Y=Cp!5|Nc
zpM&$8-J6v2-%7{&Tan}f#9<hAQA&cBnK-3RB^S(R-`LUz3xd=A=Zzg2(Z|-4+y@kY
zh}=2WGl^R9B!GbXf+viZy}7O-$0hl<I4;}o+2efrQ`wFdn81zBP#fWCZDxI_q;HVx
zstQrxy(c!tKWS|+oMm+U7qIb(<qwmOK!a}vQUL<=RqKV#cJA4u2EOuzczGmz!Na3~
zezI2!s?$ML$7&jTjMnYFna1$@)N#k8CR^$B^sX?lSo@^wCK9_GcsW@>KU+TV*|TSD
zot-gwA&b_+3-tPg{T<Hgrs}~u?m0ERM1hb^(bJNbLpT)Kx|f!$)B*wGRzReLG)#b7
zo6gLfoOm=)sFB=6UXC$$+|pH3qXCVFcgK#>q`7d90ipdRD0G60Zn2B0p=&=QW<0PA
zsbZqgg5h#0z)NuMoSK><KL0<17TRWdKgR9)`1RQN0l<qS6a^G*ZM86Wuh>M@**X1&
zF_!}AuiDAs(Z?QbRUm%v_&_tcq;Bl%)QuAn30hW3=|1pRKf@69uU&@L_yUu=DcMjO
z9-J)Zy10F-FEmV>0>n-*t)em8;tcY#_4Uwt94Cj26$qu|`esHesC4}J$9;U0Ct&la
z!?x&mX?wpq@((*@0LaCCnFirtVQ05~mSbgE6?G#pFtD!yY%>P?L@5xE5nfp`yvP?|
z)PYejY0qM%07DuyXPiJHa&kQ8S)hlsZpIKzx2#npuK$gsq~5H~@M-`bLNxoo7Z`qc
zz+VlMCVtD|tMb{7y)K|oME83Z8A;+Q@a*CcyXXNszEDFZ!FxPg{zQvMPj$?lHEY%o
ztV9wM{ev~J_s9)I^ezH_i>)R~rY@n=m{4QzH5+fE==_1eVgPRV`1y}7->~;HLU$+W
zAjuuPN^4|oose$*?HLeIdxQBj_bLUF=mEubt)0UbC!)>Dl_KH-Sii)=e*i2GJTH=v
zQC7a)T<JK8K5?)GOY5^Rmw{TzvdTY$VurTKML8q0MugZKN%+CtO8*0NI+z0hY|5mO
zo;`au>ES_Wvt}lw|FaqO@wZA0WFAJ!tqLT*ZyY;~b~CD)PAHwIPL$u`ZI#sgy;{qU
zW0FSrUUY>ijs2HrvJA_YYhos7=DMD1cBqup&2>$Q9l+Rv>Zw-<?ED#dno-bBA_E|3
zHjOR?626&G8TWy#&SOk?3xBATm_x_emRvXY;j$ezHt2I&Fb*-@l_@}J`^^9zoD$B*
zaQUBV$?$+NRq>xrBu_EHO25YG9USik7>6X+tG~@3O?~@uH89vT*o$ahM{rjON<~vn
zW+QV%#TAQ-d>%=;)P=dUkUV8KKxr-c8n4my<tXQ0-nxfwrvl0#8IHe6oCL1AZS60{
zL%96Hly-~7L+OpS*i~5_LU;{;{)!Jo_S?Pl1kG-7@#(l36I0WD=%7-Jm=3K*!^{*&
zS!jGe5-eW;u1CL_M|7hv)**Wm`0&qc4NftqK#1ZnDtfWBU*NVAHy)NggpyA+s32<h
z=1Q8~@B@i$1GzHIWzc|#pd^WhVQTSA3XfrX5TjRE9dn`%Fp+GZ_74vYe?K7C#;z1j
z+X});mX3bO5<Zy7P}{T*4B-8L{P*gI&<An{_K-iY-{l5A91w7|9N!!~^h0Sg{;T{a
zmezWDsqg>vO=~m}SxCGIELkI7iyx&9t=i_5#CAZt69;}@20qq=#7h((_=|#3el*m>
zc8m3F)Ta+nwg{9)R7<eC>f#_2zbfb>ny6wxhwj*z@Tw<=E{*yCSrXU6#AK`0HCkV~
z%~X*wKVFpVa)cWM6i7i_UBM9t-CZfrePSRH>slWMj<5s-I=;}#Ne<Wcs&Nw-JAwa`
zldhz(0S#!OS?)!xHVPqYw6m3$hSHXyF2BugVY<gP*7r?)o1Fj4uwOrZsP0S?&jF}|
z9HNYTn<jLkQSj^DQcfQYa4nL^Anp~RelJub5h%(l5@4l54FY@???!YK{6b*SX)`mD
z^~*-mjJR*{Xx!;1xtKYKxBw57nVFd|ND`^!R=lr*AZC9shUACR($aChldn#1(dzR=
zw{)zOoOl5;AF|u{BC#2uxf6viFbeiXJT6T))BY*k@}~%VZ~!CO-&{;F_fv$ybRh&#
zD>>0*S|?@lZYl7;6`&RYdY|AOK@a1VkjS~qD{BCN(+hx<@o~3)y8D<Vi4M3Ay;;Jt
zdi7H%e9R_6pLT$rSba_Sf_I+)kD?byw+@q^H@=ov0n7K0uu!NYUC$pHsNprnu)|o+
z*HXS)Js2m^H{Agb(qQW4P4~p7k3(}L;R(`JFwyxH2uCCG7Vu;&=Ia&BJ+?9k7i7B(
zCINT?O%&~a%cXz?P{aK81Q<pNHV=ely8WRcGYgC6drgUzA|fL2x}H9LN_-&<aw04%
zs6m@+{!PTu6w^|M>|<160rR)@g%}Ypz@F3Zx_kpfA#p&=k%k7S>tqZd1!a&{fEUIK
zmJeJja!Fgiy}PPU4NZq&ZW5CwA~|XS5vR*CNeylg5lD`g$UGpc_d`RiGmuU_Nkc<(
zMT3KCeYI7LF4Lr#fz&q|+WRy?Zf;pvHN5)Ssroq(-!SWw8qEJgY;(2RKGvEA&x`eX
zoJ3``_4LHyT4Mol=}*OuxS_??=DIl$l0dRpvm=#hUNg@8r!rO0p%!MoFYLAYx=S-r
zR})ZD9V~DKyXX_xRO_6(sn$_H#0HQ{zPge#37lggMno!6t9M@==SC6RP`TUe7L8Ky
zO|cRq8hjNw2i`hzgBHfWdiAR9*Dv)#O|ck!t3+}q(5R6{j|E@1{84npvLQ8=i$od9
zD=KP$@`+W6U-9ztmXB+RP)~-Kv#3m~IsxFH9!Pg-X(%@m6Ah7YNg@K=-wgC?Ul=`>
zO_{&{9sK!UZ!dJ{3Lhi70X&`>xEw(O`nod{gW9-T064Z21<Ttapz|6`1RuJz_*){~
zQp%nQDhmM$9^-u}R?vkb;<OUuSPfhcA#Dyjrxx?ECi3V?1yU6EfenCbS>?(y73xBa
zNzNWUoy294@gg@F%5x3S7gEM4MFKF1u4%Q=8bN4wko^8<OCNi*Cw&|cTNhMW4M}>E
z$@>y{lyC3ey>*P+?5**nB%}|l!}xLovU?o3xR&cMnACxAiBU{|gjS+1&z?PjkI0B<
z-bynBe4ExdJ~Z?hHaQS_Xb}$}OW_5UZ%!1(o4vz7ZMow?COy*=BTd~?b;pkW=r)_C
zp=NhfP~yG>=wSn*3et%92Ym9!A(iL=5h4;M$zvi!jF653j6y!roe!#(Fg>Be+_*-F
z;gzdb(Hitfs+PpX5KaP#mJcKjAcl?=bP-a!QZs$R%fSFJp4CyQ3{wmhrdO9yzO!%f
zRFL>rlAJ8*x`fr##Hu@ucByi6an;}c1``m$12Q=<^^}Q7h+_e%Fd=0U*oh#q=-?8^
z=}`-Xz=sbXCibIdYT=4Gf%y<89*wFF0!a<VL=o>=JY2Pj)_FH5wtDn2Rq!{mN%dx+
zt{@b!8+`8sZuBtjC2ndWmVwm-qQ{e+TPg8iNDSQE+$8?$+)qFa>L~T>m#w74Htx6n
zjAc*5wgpvUR{Q%mcL8`TfOK{9J+sOm@LO3)J9gb&CuooWzSCevAFml74U}Vv0&#B-
z(q@Q14K%W7HO1juYY8v;UaIli#bk`35}ju%Bcufthy-+Wzg<rMe!MqkoiN~tuZ2pR
zR~W%U)6H{FBPtHAT%V-+AZ*yYNk)o>gx86CNc1_fH;{3S!BjM`X-MU1>mjRom~y1Z
ztVC@S%o-0RC;dC3l3NtW4nXXc#0CD71**Yj-3$8<qVhBQn*|p)o~)nV1KT1d++Fd_
zn>Uc84%28O-|z5q-AlT9#I>C6#6oz4$eKv|V*8-&Y2)4xgjJo~^*|<$nW8$6P?yhf
zdJbJ^9>=N^Ndv3?0uDktQEy4B`ts!q2A1lfhp3k7pX+uHHwUdtmjZMm$5Oyt%S%gR
zUMkX2g*>aNfaIj~mt$xvni;nG1|~GrlU4SNY*RgojCwwp7!BD@@sJqQuLsdkt+`8m
z--~GL)w_d+5r-8vfR)Sa-7SYn-WNML;fWGGRT#U9GHx1Mm%d5V{s~E^*NsEux0s#$
zP4b)!uQU*4SVQyjZkq1U%>eKH<H?p$oNKo$hrS@^Wr%WXV{E0sd-7G(fddE77#)gP
zrAd_UiY(yqCYn_x`uSI82u(;k0i_q?jNd<SXeHWRH8=u^U?Srbk0TQ#InS$=a8_bF
z^iu$XT9Vj6M2jpHaS)pmJ*k6VWN|A+UcF4a;b8Azd-~}B;)_-&KSEn`b90$7plRq9
zpT!2lQ|W9mq@gw^FZK1km(?QE4D0qQ^uH*?O7tP`2XN*6F4>rkxjP0eA^zlRvn#6m
zAs&6DSrpn-z}llMFRy~Q6^NImL<tVTfKsa>>;4pMY4=TPDf1HmtQU}M(4bQYEJY>f
z%wYr(zl@B=NV(<M1IkJd1wLlDC#9-Ppjz8|dUz2n$}#-_*AT;l@4F=p#msI?Lr33*
zj2ZfJ2mub5$88PL&>4{{J8qZz`|3dcjIGVLSF(^;?!%kS_uSX6$mnE<(@#PMbkvhp
zAv#<SppX@`Uu5u`StgLy)YH#)RE6_LPEEmoGl<hoivzRsuRHJdbO4G#eO_}fqWDPu
z6S<dT0Ruc9-Mge0E#v_0=GKG2F)^SbyqM1DCM1&%g69MqZsW$cmT9~P4xGX;se>hW
zJ~2<{AHHabJ(ivhZjWq;U|pRaG`F=GzQ*@suOzN?XX`Er#!WkmSTNB%APhq@&pnlK
zdSdW<G7(ci`9PjXWEg#(dDSZWx4(`B?><Cqa!}y*pFg+a9ec=C-tXnF(SLPX!D+BX
zAO14wy$FTw!*M1e<FqgU@7XLm*t+=C;I}-274YVXMnT3ruC^!RyEoM68J`X2V0Lit
z>t?4$oa*#lEb#hRy@yCkrVKf4V|0SoY=CB14d2+B@etZm#6F6qb=@Iqk?lbgFmT66
z{6GZ!31YcppuYKv^CMXcdrJ_o8Y|goM?z9sxFg&OAY8C(o)M$Ylx9e7U9+=`OC3z2
zedm52*=9;hu`^auEhb4B74L(EOksvvPy8COc0-UUk{k><bm$OK^kF(BlY?y2IR)WZ
zHHcf?>5Y90b|-P-2hB2g*Lte8&F?hj4;cd|Wn%DN%j@d2?;IGuPBezwWq>2-;ZB`l
zE<MoWv|&<=qVut;71MVPCze}%Yyws~hILMxgWEMnn=tfs1VzEOZQEh&Lv18OnAfk5
zg%EKTaS5VFHm4BCMWWl7K}vzhc0P;nnLi0%R1*!Kn96^SR;1fh!;2^3qn+?PP}BVW
zN_14yGhAvDY_qFO>!0C=7ARG62I4<?@<t&L4;275?Bl0sRpf*V(GAG?t(4(em%KA_
zTxB?<CM6}6{=HrBU~9*z@Gz_9?6h#`ekAP#BBh3VpN8mfj_E|-w@An-7--6YQ9SC=
zFYpEF8C)dyn%x=hK@R*88=geLfc#>C=~??^S%m5yPtC_XG;miy<Ij@1v4r{*B`-zC
zwsx&)#vFn!5Yn}Y;Dmca3>(}nLnQpQfFiid(wmBrUUT|r-^P?~_dN;SE*__Q;tUJ&
z5u+)dwslj1)Q8~gPFUedlA@wI;L531R`oA`?2C9af@l)B?C|qXdf&zjs<Nzcccwv|
z+dDh8eQRlafAItg`z|=Uz0t+oQmFa|wx4ln+=W0R(QU1*t)JF(?IeW>b>=LV1M%6=
zaomO-;jS`J&93O{2DEo{L<9G5h}st>Ndlm|G&^Z;(QH3ud%8mcn;y!H`>$IfI3`!W
z(C?AA{v&!MiMkRoq&eGZ4AFC5DXILsar7$8V&et$c4VlNkXWjPI)YE?RnL9C<YZ;V
zI(An%mrf(Z&duHKHwO;baODZ^f4kwAD}QK*v^Fpim<mZ}le2X(5$p3F0B({jO1jt7
zH%TPNNs<rEc7v2>=0MJ%6v;P|bSHXwSn~JZ#kW3`m3`^@5=p$ZMv*%l-^R@LUV<Hq
z?M1(MaSt~5G{IjufClLx;F~8U67_ybp?QRt7O_N|wHcETS}_>Yi~dhsrF)n?9Blw|
z46*PxMFPc7^}~7ctP%DIjDj?J7QQ4g{>V8ga%_rV8;ukb*^globzQWdwl~G0J6E>v
zW11Q5R<D2FbhnpZ!f|SS)dvI5-?Cn#Uydp&Dv}uZNX`K&!*?qO1<(?QWjl+ml07q!
zW77}oD;v%}P8;-@(WtAdCr=BO;wYsB8vGx<vwix=*TkI=+t9P_VI&wo{P{U)qA#3_
z^hAhyf=~SHBhl?a2gyM<!!c><<}0El0-TarjZvdp-%jn^icOfRo47&T@dea9o8II(
zqPO!QFHHgkNJ!k}Ld!n3Y}v99DX(dk8AlhFg-Syk`y+MPny?#U3FXF2A08X4C1)%|
zT3ipqp7=C&Qn*#?H(}iPsTOvJwGECRLSI!uLJb{6<9cm#F8H!8Hu-1JF=Dep^J#SW
zOxa7y?x9w!-*be(HKK%|(Syk%%}UG{^!YK!r;RPPV~C?*#SyWv^x&J+a!S+-Lbl))
zIvq0y2Z7~~4UR%atw}if+Od?wg;}fYU_Ln<cHT?&ImlU0Q`Tz?TeCXvJUc0$srH8i
z!V*~+0h+{<ujO#p1n>4vW`G@8LP9|%r5Zv5!O&h1?+(x{ZN17|Mu|ePRv_9%UxT3Z
zQho*ExRDca;8Vc8%0$>WeY%Z*fxo)9InLkTe^;i7k-Eh>ePNC(F^{`D*E!@pbex})
z2}ULkM4r@8&UkO0=Q;Rw%|NlRo%0~aM@X|eT*B$hr#OQ{4gwJc8!3e6a5+^f4~!gp
z`j!64*`0wjpHajUY>0sqiw@=?kFOtsyS;f)4v|R{!^Jp;#;dSmN^=XqENX$-d3{E1
z@5eQINTY>I>LJsq0lk_;oQH;nw*DR$3snkW@}-7OhztjQk`|kG@Tm*=GK$koggKyo
z%Fz69R+q$IN&XCyvP?pQ*eSaQlSF30@uhtL(e(86-6XpT-jwD&=e{ZJ!eJa5^}Uq!
z)xQHWAUVZJrUEP)6-0hWU<L}(?hhJN(mo~YG+h=<9H5TIubN>}UfG1G5!87{a$M)o
z1BO#;;peeeV6P)O+E1IXSfmH-N8Echlb|*$2NxG@93Ws}Wj#t=Ca+r8n!vb&ZgLP(
ziJ(Fi#7~PlZ!!$;NY7_>WRVO5SW=qV#|`ATEV9rf9zv!Q&VO>}H!v(Y43)O<iI1SD
zcVe=N{x}+uB@Vt}V`P)_N{G9h?W<2AyUWIjv5<?lmH*-yIp5rzmh*ddoCvQ03x-q7
z+r5zH(s^!Q%-lkuVT>H(7Nu9*>-}H6Z;;ul6cS+Kuj^1?@HY#vI^#x@knn`d^(IrL
z^)jk#BVpF2_r1PP<UHN^eV1&&5jEBAyixl41=86wv+O$@l|ML1#Y~L+7QJ({bJ{*v
zO6<eTsR%0?xE3^&ySw}6>gr#CC#{hYJ7r;45At=SfWut2(olA#!4<%XM+R>uir>8}
z{PbxB2q+B&QE3w{`v|JJpMLw;hB3aiGVgt$*}VDkMYZ0_1;EJO(eWL&5JjPRldo|A
zpN7I_zZe<04KC?N{@3;FQl4ypUo5tmpVX5z!oka#n4ZMMT*KLtr-_-FY&aZxc(8?Y
zA&{JaK9%RLIjMGTutgHbMGl+g>B(3!j8SW3j>JcooV_~#+H+;|iR1ZfBDU<1wzan{
zn0bxOD|16=l|ld9ZB$4p-)I|&H84T<T0~F3?55)qdiq0pdK)3p%y<3z6f8uSm6gRN
zWU{EHwpLYF*AHU6=kVa_XFGYRWm_)W1K#_?y*BQtiH%_27PCt#tqabT1Wrl)=<ny>
zwI~lsffGqIu01t82pKcs2#+k%o2xiER~i}`cJ=m}V)fJ9m7~0$(cWDpbN|TkiL<;p
z*k`TKmY(%!&S?my9NGE??Pxb7sjN*dU@1U*i}R;@_^3i~(F@Q4tudY$4yy5;W!k*u
z@+twSRn*3f8}Z~Lw~O?Rjh|MW7qqmr+{wrH>;Cl@;ypB6mshEEhD&+*AsRe8xG*m`
zxKq&>66+*54UQ-2{a7;uEpNB?tZU1c>weER9h6zR#c2Sc1#sIdFZ&dnus@$dz9YGw
zTL%pE^<JG`26#dbRN>Dqwgx@n9bl>uVoga|@8&(YIprjMFRzm6(KvZ>{R$58Mds$_
zp5rISq_09HN`VaN*t`99%AFt(+BZ;cQ(S+Ww5TqM8zO&5zj;ekOS~Xs>|jG0BR<EV
z1mZJAfuy`yn5|qe!kl#q!C^C6jFo-o-Gl`NSE7SmNKO{iS57)qJD11%!QqxCn?CLK
z%Xi(0nG`Hyw+VK*A6B=hs%i~{*&|3qg>|s)J-cd?l;_fI+afKcAKbaK(%Qzx*v3Ym
zu1x+)&%8lUdzPG(zdKeuD*WZ^*Y~oeT?S6|wd8F?jz|XODE3^f39Rzm#DG3KPWvd&
z=S{ylcH;P~fo6YJwBHMg3orj7HI*4M)FE$gsfv#u-AV)W+}*_?t6v4}IjpXJ&xJim
zbz^~cigcX@3k&X-+MT8XA5QR?qpGT_(ZsGp<pwIc;<78N89)<9B<#<hU(<jYb8c=L
zf@%sK&X1^QYj*~#gjx<y#PL?{)0D;iyIivkC*wuEXGP%!kQq_&SY#z`<+|o2ub)5X
zMlOoKb{tqd67Y(Y7K;x<T-o>dR|oLEa;^eJx8lh5`-+MqDk`hMh74lX-(UUD^}#87
zg4(AjU=18|-U2x4*qhtk)%D(=yZY-_IRbFK{U7bA6i@!=31I-j*z?B>-v9&BG;F@H
z+@Ar}6<f+?x>G=)gX2;SU2eD){NSXNc5|E-zX9W5tFZ7|vfwC<FUf`l8{GP`Nm*R3
zSqm-w4XW}>^zn^t?d?AxZ(R5DD?;hI9*tj(ci6#kh?{5FL$kQE0U*A$x0ey;chPT_
zAX<6l@#Bs7$by@A^ziK+B4jg1g(Js653msM=+O-{i%77V`gEI(i&wDg@GTg4o!o!r
zeOzy+uKtVq$8I>oL;1pZ`UN$tKFx|1M7mTgZ1$g}Ho-2<-FDYqrZ(IP%`elTa|I-*
zEwxh6(~lfI>Y%78nYwl<_NvVLqd2Yi1eJmNeBaXU%QJfT_HTGJ$5d$K$pfBkzHy{T
z;-XB~n%(5#8d?Shn=m&o&tQgLzjNnWQj%@Z4m(gVrk<XjJ$YWJ3T#r@TOEmfHFSTT
zAqb;v$<d=nMa9KwDexj1e4T)ZPWc@)V4AgS*W%EZY}_1RN>_8v)vrzorhoqMNG#7t
zQJIOBp_zrAj!r9rQnlWpIQjUdx_WJNbo6c!ky6kdSY9-&tPeNHpPV4?e3i`7smlOV
zYv7zug2&#)jYJWdfZCj`OX3(j%kY$RISr<<;DBt6h;TdX<`OKMTj`6{dl%z3OR)%l
z9zwoQ?WP?9Il2}QusSt0^)u=9`JRtlxbUdUs|GaEsiq7&PMkTug*?QT?b{cFycNMk
zx*HNwF0p~!t)<$$63e)D?_LF;AXv&voH7Cf5p@}0hhP2@9|7>Gn>>>mjwC8<JUI0*
zEtZG(B(;u}^%oX17@SUH2BEK2>(`&`{?5b8E2F9!V82isKm1J|Wtf!rgevzNwrV?P
zprmu_z=}Axd`BtH>xVi=E-ZOA;>H?Qxy0o2lxOROE>GT?nk?!duHn`JWt+N{EO#n{
zFBbSjUqlBkr(R;H#X4gae1-6%xNtioLxk?2NcNXowvLe-o4dA$=pjcclNP8rPHDON
zTrw-EUm{{-@Y09|fsAcUIGJ|X*?B)s6g?f?C|Te=w_)+(#qvr@hao1pmd5TE=WLhj
zjBM)iyuUH*&iS>~Q@cy<K7Op%KIi{ECXn6;X!&VU(rwsZ3TftMW-^9`1yu!+v9ZgJ
zA3q+uUthmj_k+$x=xEPIPw2RPp4ut$-{*U*^C<@9z_6en6iT1)^yvj4{Z_y?wSEM$
ze1n2mVl>DKY$-p)?LpBP!7H0Am7c!<yhr*jXqcR(C9m11(6YZv)dNy4m3z2mF-j5w
zGA%xCpb&P2ii(Oa$ix#|V{T+EgOig}eM7^~*OeNP(J>>cwvJNew>}-s`hsiIl!#av
zusgbI*7#3>Y?b2``=_KXd$N&!g`bWw{V!k0^FzZah0}ee0hse9H0>2PmZA`MJ}oMe
zX<5xY3?k0^eW@W5Y&&^*11$sv1mxu9mtH$^Bxw2iJ&Qn7_Vg3;5-O=ES6n!5%oY(5
z?P!qDl@OtR-F@NSV|79KL*eyf;MhS636CC~<li295^BuQ*jO=o&2JR?HX)(au*)t!
ze8`HR!%~=WEvu#Ru{(+7Gjb0fJotJ(B;+|(3nUkZR}|wU`S=(?6`qa!*u}-Q2(iMu
z-#)zpIzn?R#$f?bZuJTXH|G#DdDJfo?y`Kxjvcee7K8@!$Q?hv9Ab^g=;&xv1Mc@1
zaNI9mJhPvBJJHh4VwXHGzQ0F3ZQ@d#TEFHBO%=Si%>nX-Y%`}*wnXpfN6%Z$&d!9$
z^Xy2(0&@e(B?Z|!Rc4auW%l--ofN{cgq46)I=Z^FRFXlSAA<S`r?LXIk6iKU@-V?f
zgSXQodVI{QYitA^z_LHF>;r^uV>>&+5t;{&AKygB_DiC29lO`yB;J_@Mn=YSQ^RLg
z++Y?)V3YDavh@NE)@xia03<^cOV#GY^z@k3?vg^b!?v~p=ognwk*OB-nog~7P+^i9
z^}ZVvbOj*goV<p}3yS#>@{2z@aa!Ef-Q5J)3N=}|g9oYExj8pk4SB5We!CVX2?>eV
z{^^Q}3LKy*6gpYEeA5A)$tlDV4jCD3hL*iERZLvGsHTP))m#GKvUJ<CCK8~f+BuCd
zI<DNlecK)<E0`gPbbU?SnCyC9SVZI+E*$?03k$1JT)-8|D=S~k%abg<?57%ZalDz2
z9Yfr8yzh~H0f7sZ;j&PQT#u;B7Y~LDT}7cB*3;7yjYj9j4T%-ZEmy`(cAaxXPNO3e
zKoOrdBgUofAIH)){ESxRG}H3sS0dng+%fc-V_C6ch0NnekJd9WF}1e0)1w?Vt!5d9
z*x)$ObTT?GA%O`-4*MF-O0d|JbHC&sjbpc(K(%0B!()bHgjBh7;eOkurPzZ%g0Eg(
zj3U+PcH@ort9yqrz59HdeC);apPsnLQoisejl?I3nMGRy<aaxEsXY4C(9lr-(wOer
zX}>984H^o2`Z7)#b$Q$qFvwk6!iE@vA7t03u-%G;mM>q<wqXMi_}JH$AFX9@bW^3-
zX@XBx*$sowLS)Y#8Bfmx*N*u0eu-QnV3@y4vy~Q~SFps$*7g%m&`-R-m$>{zix$bJ
z*)^mtLrV-*D@%S>>W0L;sORKvw3QMu)iGBO54xQT2j?Mco++ugaVw4`6+tT_|EB;#
zna=$oHNj%I5VF^Ll&<T{!WMq8K`tOL@Zz?kx1V5_5mHckdN?+-JRmgGAHY__t>X17
zwKQ6NTmS11va-Y+&Yd#`RF=^*Hr_(M?&#=XzgIR012F*|!WtkU0bw<3oWQWgT|RdB
z@TD+mpKzB0S?B_O81O0hAP5Xv5&T^VJeD`8>UwMo*V6d(47=~IP`wYr!ahp)$v_o9
zi7{a>&CBZt+2JNEtQ)4h#Wvsb!*Gmv`1T<wBcwrh?c8}?s6<Cud4*MN91G49OIBA^
znc$p;M5p7hghK~Yj_1@X`voZ7rl6&#;Ttfwd&mH;A~U&mK-zmY7$0@;01Cxz)3}DC
z^iQrM3!VZ>p8^oO2b;yr+`JSS`7(4B9SsfULD%yV9^<=Sr=%Fe*fc=_R5Uc`)(93t
z=}$+Dp`)*_A~ZC#9z)|KX2GBi!=rsx>l(A2_)HUWCvj@nWX+AsNY`H1&vgy36QyTj
z7rhIqes>AmhxhM|;4zpzc<>+rv)>vTKxlQ@2^Pi%;0cC_PCzkm`SM--{NDhd#kQ*-
zgVs`y3uiO+ha++YN@hw`m0CJZmEofiL_T%5nIj_O8F1^?tuq@oY_NvcV2wjg6}7cN
z4&qa>x)%&IH7LJRfL@KM7Hda_hK!J#Hp8wrgB9_ELrDN<5zm0tt!Wo2ohDV=hjHOA
zR03_g2+H(_f$s&CIMe$sAt6CD=3Ycnl7N!Dypg|9I<$=+?C`EjA@aD&sb5D0LKiy&
zV1~cOhowA9Pv1{oa-*dCUI%CAj!<Q9+{Rtp+*gE3>QP3IKvZzE?CZXM?O1<&3%Sjx
zVEiqPnUJ&9c)!iLZuh!-dmWBNvuW~uj`u5}{4^94mLVqn13dso5L(W~$?EHKpr+sD
zc}yIUE=eegS+_DRB_%eJYu`39WAri>Q=w8`8weEvC-~9PvAu$M-MzrTHO8~wedb3v
z|3FJ{`l}dC;a`}~cYfj+9sJHoy~GU;4hh)`eue;3Da81c(`9QuEo;B`;KBA|3=;06
zJM-tKt-zPxHZ*KVPfz!J%ku=hQWU=k4GTLr&?M&S>WTwx!ov7av$a^K7yymgSx>4x
zIl}J@aZm)Z_bsQ{EG^B6dcajm`)mRsxe^kx0iOVJH!!eBOuzqW*uf)l$Tf(bZ(BOZ
zwyy*Y8{x5^9FZkJ66H<MYPL4LlQT5rMCUD1V~1u=r!rr{Q*vkuN@gLt8JRn$5KRf@
zc^n>o8w+W#yhB=(@2r!fBZdn7lUcf~pHtD#pI^Cnkrr?N72YGJ$}?OD8;<3@w624*
zBP&*BgE(xku=xr?`5*)Qz*(0?D)}#+@L4WmG4~8*FN8mMa24-<*;<Epx=*yh0k>j{
z_=Bkq-&Wm*#O{yo?i(m@)WzXZy%?A`^3Aw2gwc)m)~&=H?&LUOi{4d?U<4hex;)n!
zWQ*CURV*yraZVK9<kS?&@4ie+D+WFPf&-^iA^gQF;d_Nt(=IxrLlBUR55Kz&hxut?
zrpJT;Ia4SSN=sG=pRM;3H$xPkMDf|y+<eEbG_LIhbK@IHfPwvh>niw7MGgKSxT|Yl
z9cF3a|BQf%;52Ps+*v-+)gwST=(<Mc=GTA|-ybt3hk-@8??r?(EG!5N4^+|i6)-mW
zS!mz8vux!`#Q2xsv~HP%U(<?B2d+a~^@pi8`GAw1-5(L2Ws=G!Y8d-C8K$&c!WCeH
zoLS0lW`f90w1r&2V5IXoK<yS&I+^xcAZgnJ2BGjqhTD~P8RW8|xlST(dF5GAW8+3_
zd7ad=t2;06yCifKxAEq{RJ32Xqz5VIQH0qh58wlXHUI{1Ef>tvJcP!lnZDSF{>G85
zaY}JGp3EsZKOnHt1mcw>S%^P9;khC^JNrv~z>V=64S@Suc-cjp62Z<OG~Dg_8n&oh
z#JvF62Fi2m&z~1@W4B34auA1E_r!@^8nI6_Ss+6nQB?GM8U7UbV19Pw1gxFK6k$Y>
zXfe-bFFwA1|1wTUuf+iYodR!3XuDU?mx>AtDHLui$B8W~Ii-%MsjUIze3F=`|D0dD
z7;O6)Gp2<x5X^{)ss9#!{^f8?TJIkd8nm^bzaMc_PHr(;`W0;MBS((dD{haN5K*wJ
zidqJ<{R+%5npJ~KbQBdC`uh6=LPGqWJ==;;<a|by1Dth)=q39)2Zw(<j^50{gxU&?
z_Wr{(LBo81BpZJ~%M`V5rI13ohUs7U{{2$uxNRKCiZwMgLw(<be=NL!TaoEClU~ys
zgFhgG)p~VX++ntTmdpu>iK{nlx{bL@=JU|RL@5Z9JwAV@Wp#TXupLD7#Xz5z0fL|(
zAcnFdGb`&5(2?08AI(Fil;)zD!NtucXHK8Kj5Yp{ibC;8jZr^|>C1=#+lBny(CbT=
z=zun1nfVOe);=>R9rJFM-J=fb+h-tTFm-vGg*NO-_+Pin+OU^4!kA*1z}&#i;BBfb
zFJDt4e4`9!tOy!9Yh|@fTzmtjB<0)GbPD08iw@Z#+$Ec$(GW}_0%zL)pal%RJPz-^
z6|M#QJ)V-cZ|M*<I*5TtiaPZKtZd7Q=(wk}Ds4DvU}R~@U}Iy0XMS5z!GsUHY2{F|
z<?ONa8R=LKbU?TVfb9YI(WRIJfR=JtDjX7D4MF_XFNJq)VEH--4Axv@Rmra#hu<Og
zJn;Jr{S?~D5Yps#=GO~B)Dr{ByADAD;h@DbIpA_|dBnDFJ2<sJ>eHuBd-|aZ7xmc)
ziTTVQKveIv*=`pEs-Hi+@YIQt-37Xhv;C8G1@rV)RsiIyR<FK*&kcoo!&;=PUTzI0
zNR=1Rd*EP^*9<Q|zbv2va+$?|j7S{b4GX)4?8Y)1Fs-5SaXHLfU<7#Si@)?8;tP;J
z&EOz3u9KULQCjuk>xu9KtH08zvzW{-WMkGmw>q=mqG}Q9t?jP7&|rZ#uZQG~0F2u`
zd-j}8`t<JIyQRVxi@5{^PxQQwDT3#EWW*qGDK8~UjMAkH3~Sc?SmY~YTJi*D^w96$
zt7U26+0s49eYh2g+_<B1hY!;M7y-8K0G5Z@dLc0}QH3gjy8ZFvavZzey?=iRmXS!P
zlarG}%XMY_g)+0V8SF0Nb76wv9Pj>>`?D&vo^)kMB5xrx(A@XgUEnsdxTU4V-rl}7
z^fuY#*thpj4#RZgE1DD&0wBe>kohvv@o{7D<HwAzj}_NTFj5F)O~OS#hcFBpJ@S`Z
zFlJ$RsYM($F*Y^5gmYHA^wOEA%MK&~wX@D9kr(fD*S2kRT3T9r`th%4Cuhg7EL3^j
zBfSFeye`(}T&La#L)&R7LLq%==w6=ru#WP*aDfJJLROvfMkYjU0`mT@5%YJGR|~-S
zm_UAE93BP0h2#%4mM+!$+ZtpBe{dw1bFEnh#qlXLMy5#Dk&at0ouk?6J}>fxw`>Fl
zok(kV2tJ?X+*pq$CZ?aI_C55odoGuCe?Y6l)=jmlTj4L{IM~7greuHa9MJ&XO7At|
zCL3E>-9+N@2?+1P{6qm!UQe5w-!v-<-+Ff4%j=JVgA0p`=^>=ww=_gWqqwi(^rCWE
zk03uk{{hz_3ao%C*RH^azp=J3b4q(NLBwq9=ot1_pb#lGLNfmP?b~uttdWbpg}|D^
zaRyaq$ml^yaQD4ZK2|OE1@<%sFT#gyEuzj^S{C60(s~%z{ov!_S%!~zK>NA`5|HiK
zT_h1$MCJ(u+@@SNbyw+0beJpL9vaY5w>skpBmyfmD8Wl<Y5Q<T5C^%44%3Ij{Z)su
z%XuLp;250D*|R$_?$11sUG*X4D*i#va0yzw%BfRK_-v3+(Q~xU&dyZ1;IfsQ4zNLy
z2)(yPb0~I9nD4q1$47vsccrDs?J!}vSlg%l2WNg<;~WA8d4iKVxsi9>Z!;VW$Z;Fl
zj^t{<-MgC|i!1TLU5<@)$3jI{qXqv_Z(zc{XS)A?dOP!YtkbrQ-_nA}E-H)(NeeeE
zme8O^H=`j+%`<Ot_ozH%ZD_Qp6e7%2V%o?SYC>b$Of?m%Y0-psL#C7#(Z0XmOY^+%
ze4gi>=g;?#J}pYW>-t^Sc^>C+9Orp&1Ev?x?aqNKW*|&UdwUWX<zqWLeMtCYmuo?=
zfiMXBm%VjpFL?TehsViJGg7-M$7v=#zx1n4K9$eXCvi32V}c0h?=G}${LI6*kDo-V
zFt@BMmB{|{)vMpdZU|ZY?u|X2qN&hhckbNz2x_@>SQ9dno}M0*$*t^fs}FtASHgG;
z0shF~^Dmd@?Tnv)1S@Rljf4ITo4yCwbzitrxtxmJu&88ar&ZrklRIaBX*qoO&y)g>
zKmwQr{9?XPKYU1c!RpxU%Y%lv`sj{Pc28_Jem>2B^uE_9+Xjx=0GxTwadH*QYlZoU
zd)lEJ6BCn5Tv_ynvePos!1I1oRMgz7Mib%^kgkmo<e_|SCnhE3=jJBFriCnCez%y6
zdi~&{{;?0Qf0L8#-qDP^W$sn_P#+<9Cs&p)eg8JMprAMB=23UI?_H~V4T=`7==85!
zDo|-cFPAH%Nqv?o-M@Z7n?H3USQOrqYWfabczyp0l!u`vCM#lfR+qS{UrZwQw9vbY
zyGpIoqCg9yR1$$|{`9;T8z8no)GY*_>ih<k4h5Y&!rjRquoljt=&_5u*_$RY(}Out
z*seWfvXzGqI|BhNs$RKBX1(3spiJpsP+a^t-DeaB&}NUZzZAn6Sbw%opWF$akCK<Z
zEK>oF>^Ru{WH?!ZxgE}(`c-Q<TAov<Cc$lDdXthG#ZH);n^((Eojr9bCAPP3hM4-+
z;Vf6h93+@ycb5TADU7OzHhhN-ZEf&BJYXX&J71anC=J$iwDaE?_g0L)-mvM8TrMws
zIAhCs{Fn7WTJMq&V{R2$eFO=&7Z%iskXIUm2CXHTb)^?tBT_YMdN_5z8#o}vy%*D}
zI0ZT^7mA?$t@`c^sjY2Ujp^8M1Qp-sBb+nCukTmDOX6Mf`?QLaPM;rs*N!oywWCAD
z)6-LOt0pue!WskDsS!JlS$-8!np;#PCSbumEQwlEZbP6NXsVTR=G;im?T*bCMye0u
zwc^uw^!TwGDS&~}Yh_l-sTb%(j0DS=nwbT{M$zqKPtMIl$2$GiETuc1%fPp>xHPvc
zTlHC9&Gz5Sel!w;AZ0i9SXc|{M04om2w9;WRk+u7XG3B63;FG*`0>zKX_yJSlqX9h
z^qsSyZQ((8!=~%7MF9G~U<p+wp9E*!J#=D>uFJS_dW0c2fKAtnF1e7O{@^2#9Gnj<
z9BktMQ~>B%KPXYOXa~Q?eWHqrK8lg+1BgNAl*?}4P~ZJ0R?W$aodv5Eo9J!91)%{b
z%(^mM5S@681+`E-cP%rYCsBSA_<-3eMB(OdlDFK^Dp&d}$ENVK-19^bkRf8xqSZ%^
zxZp<``9lUaX|mst^_u?QZr!TJdzi5Fl@5&e%8Cl9zI=adrxD(%bV8{6<jJRZY41$?
z4>XwB!CvLLxuT0W)6yE-hIwh&tLDGYnUg?4d0N%#GG4Rtu!4ZU9r{WQ1|P{~;OHEg
zsTQrYigf+<qtLIj&kMqc&HR!<*AK<2Yo+wAt*u38+_p1m`ckc-?#g2p{F>;r<VB{m
zHak6ifTv6rlqP+~{uY&#oU$F{V}eW;#AU!nFwdqAsFGb$GJG>|I%#fWeqmt^v*7-I
zwXLl!Co4-2&I{0|&jeVxR*HK^QE_n{<Iv7IZgT~j*MI;6GVBkUK9xu`!o+6;7&!qL
z21iFn%c%$$NHOa^6l-;x{}qxz4$p!W)A$<4BsC=^5cM^b>V6ko#x7v^eMG-%`ROo%
z4J~TUa{A@eKnYF|n})A%Zw?PYZMlrL0~s||MDO}i{}FnIhG+HHkCu=*>X;jHUM!9{
zai5hLPgur??cn+I17cT%B>yEhHtge%SJDF~Hb0#aHI>YQ$XdVtI1IvN(p?UuV`^I3
z?}pP%XmSuV&6z7#a?qM8u(DS|D>YaA-VMWT3$_Q}m`8&W)5Mpla-%(_JCT*-n8s}$
za43<$77o#BR2!t|v0){G{Wh27=?sF)hks07w{G>pgAU-%Ip8Qn<F0fMzl5S9bBeWk
zC{bAkHrw9hkjdml!(nCHfM<am2ok<_)^6&TahzrEZ4*-HPmA^r7z^)eZEby&-a$(z
zXL+q^L1AGlD&(RqX_8y_Zf>=Y?u$oY_mdYW{;e%72~;PSA}c#CG&G#xH8BBnUdc1Y
zqa33skLm2^KS?f0@fES10#j_d04o45IAencn>|BBa~GGlO<hlPjS4N$u(<RvPcqcr
z-T<?pyQ^z2eR;NZU@rh$Jwte!yGrj9rU*!{Te|SVYYO%Cxck_l6DGXLuL}ze?XZ~g
z6kvon^nO(xyS9$e)ec^Mw$QAERm*7-Uh&$6X+;xI0A+a~C8leu2)7JMrvMzVO0@v^
zxKD>#o@j11f{KIj8<?AyXR>J5&YemW3+4OvwQum2!ahgD#wH<uz)mf$E6L(kWjtJq
zrEwaLw8p?4-vRnYTzvM>-MM1q#fug4Pq~#3R!#C@7DdDFrWM(av7^OgH<4U;{h1!7
z;z>j(U}qnv&xp61z0&;Y-@s9}<HiM(6~489MmTGt2~E$)=$wgx&m?#Oui3M=GnQfn
z9mBM?r1$)w!NNvSQL%~bzOe3>ptK6!rAk|H4cb5&wfKWob?=${5gZaSD=8(UmRMf6
z5aaJW-sC*RzN&NHzdXO3ds2yix*VFg93Yw_r#<*geSPMrnU#rdl^u@I(tCgWF_^qP
zPyA?VYPye6ty<m)YFC~C!1_MsQP^r0VmoYDv4fYp-wpmO8)krjsU4hMT2fL0qNrqS
z#iR6M4Y0GGhQ?7q3r@yL()|nv+h@ouf-W^)dPHM0ZEj7>u4tB)?gFB~OGruW!QtWO
z@#AMhzYMmIGMzSK#yluiU+r&_qw=SGlsLl8&25WQydL3gqtR3kkE5ugh^@2gP9>LO
zr4(hiL?W&Q_{V;j?-sHbFkg|J#A*LD`1)^CUEZD;p0fndW6eZ}Ar3K0Vtm)5!?<BG
zc<`gAPAM~lFah?iy<lt`^70+x6Y21{qiiE8wj8W|dG9=R#!V<lK$$f#)&23U0llNV
z1DssmJW^ct^436~hXYKl#HJDNP~Owev|Ll6qDmyH;y>uHGcdUsICt)opNXKoMHlJu
zF^#09!?ONq*rv0hZEC8IgxkLEY_1ic8>*DxV`nzB^B@09f^?<(-kYzb)XMCOuaZ+!
zU(Zp{A3Qh}aNH1S>3YJ<-aaQbanM+n2(Didn>OeouE4<GZpY3H2$(}t#TMU2RsVa^
zvZ+ghN-t!1yZ`d@&kD={FNxV&2!n_!(MA{k&CE27$e>&DLuljD*~?y<kr?9Yc4qxW
zd7lD5)0IAB6gT_?QGIdUM=H6!G9#++(oh!V6QKlS!eNeh^CA)=ur<}{@I;hjK}$QD
zo5-w?5Rvuko0`_ZIdiVdnVH#KQ9264{ts>wsmz%GV1Ov2vX7&~{NP-~8gA>rTb1KF
zeG9G7&u1k)ZSA(2xG6z--m6!yn1X!*swzVa1`iwU4CZ?S;b#10+y;$az5Dh(_RBB*
z-@m$aWKLL^TCF}?Y|UJ{>bS06e6ULO&?Zl`V8K%&=Sw8$J`vH;tIEnoq^(<5L2D&*
z)b#iZOI9_xbw8Q4EHq3e<$Ss?fS?gZ1qx;`&wPM;r$=J)O6&uS++B--+oP&qzkbJY
zcs9s;f9B;SR^;>WB%dWSnCi*(vO^frQmKFc<6L)%M4UtbE7^B(?XFEwR0Vo%@ANB1
zHjJ{uK_$z_*DQ`_r+#9xH>4S<X71J9yLP#|EPL1kTrEa2DWgMWGSiA1ylt@*y3Lw3
zC}V!qv?`;$P!B3D2rBwm<dawzN`=CPa8G^sP5$rhwE_2f+!Wv3yMDlyj~K}yUdpOE
zukBOE2h+Mi=G<wuv0EA(99)%Ma2uPK2<?=!IA|uv#EfCFmP5{;Jv)gEZD+pA0bsVV
zLmD4D^)AP#5BV?KE^<fx?K^jr2pyr#)dqqq%}B1I0CZK-Jb)872y=@PG;CE>Rr#g`
z#7myxyyJgutUly-`q1jy*UuFqb@V5`Y`!QSfBSZ0VW<K+1=Z(%a;yt6cFDV!?L)Mu
zZ?$}${pC+z>r$p~SisX;K`oy%@wt&JMUCIl3x+p^8GNRurU}wDpbh!M>jzM~+6{Nh
zgu`XKJSeCI#+!sZMOXg=j7(a;UVBe=)W6k+`p{CxA_+8%R6NR<8oBHx1#lpuny*u^
zq&)ulLa|O1nw#2QpAd8U2>V=*J_uA92%~X=)NgNA_9aNhRmsPa-z6cfkYNk1GMY#Y
z{7>P&ej2gi>!&IX4i4C?Hieu&cWw<J3I-3Y-3#Cx`TAK{PHP`#;J_vhgm<$?gQrJD
z%KQVdC*|ek63Lt2-oG^lNkmZh=d|aQ<}Si|<X*r5qoRCw;o7|IDd7cENzR4Wu2Br7
zMAe3Z$N#?8MM70@;0`|nCU!pg_AWWd-TsSf1P}g0oRscdXWC4y8D$11TN>Kc?{jr*
zQpru+dgr*2Ha1!mLlU`HO#k7`KOamKhm>~7EiCLK!P9UI;wb_KTiBDl1CBIaGP3n|
z6E>@znS2Jp$bfPQZrh_6KpD_v6GMK)h9*EUS*V93d3Ik}@~_ax11GwMj*jWr+cCIR
zcHCD`Rabw1Cvx!M%?>i(eLY89H*?#z0hDCD`}LD!Qx7;0W#;AO6<6mQKyiQpY`FiN
z{rmUlLJ3yqYwG3~nL7@8cq6AhC}+`Vt1h(x*mph_^*Fw!74YS`xJH%K)HcovaSz2U
z-}3O`0A3yX(&KbD%e)Y$V8Ed7a}c-&ydGb<aJD|HpD>FidoVuo`1$kdh+VbJ85Nt@
z{7pzAH9mFN1<2%P9}lw*&y1OIadxCg_wo0tOF!IE*z{F|OwRxGyCW~EfYiP}?$<b;
zz9OVAQMy2`$T&8#j(4w}f}v=(eA~E7u-I5KV8tlWt5%Ly(B=_(>YZ!RDO2c-uV9y{
zUUqQ$Y}(!lRO5Y0>GQ+8&c4QiHvqqd&$H~Gj;NXR6wCotAselgr|9~K_pMRrdu339
zE)W#;*e_{;hkr^{ofrS&@|s6qQiTfI4Xv(r_{;}AmQy7Y|7R6Od;dxJ7iGupGwFT0
zzbjR2pqlK1gHWT;ze}@l?WTh@GXJm7{70HrBxA3c@rXY7ZDhm#@3f+n@E2)p_;)we
zC3xBkzP4?HTDpCE4FY)}t+Ile5M{PB@WWFl`zoudiUgxhNQWsirXXDf8p2VOf-caV
zJ|i67y%ovPddwIdp&^u4RD6Qtmot%$YSAn$yN~wkYn#PgNL5hmE4)xbof8HQzTbBF
zv8$8zCgAY3*vP^9$ihO2&RET>%t&0tz8<Zs>qn*<a<OsthfJkAI%oQmY=)m4yL?$6
z#zG8YZEb5iK7PI~Lfu3pp#8>rdVS;L<6D<u)5^@ssvSCOYi^IVR%Z4mQugG@lP#Ia
zlG01^qCI;u*e1!u$Ve%NFmS<w;r|TlaW!R&{s>jzt?z%x<imB01<`1tso=M#?hG)S
z&|JqriwcDHG4#EJ46gZ~z}z(YU~exa6ZaAZQ0U+rn#h?(UBRjmaqF~hB51xeX-cEt
zlWYTheI=)0a;a{5>h8Kw6bpUFuSVhpS2Z*=<i>&r-RurC)HF?Bx>9V|FGq1%`8bRZ
zZ+1suLUzNA!~RM!%TkCwPM;pmbX-O9In~B-<RDB$Y514(=A#J><aHzgU821$w{wj@
z!KN7OSi|fH=syqL)uo9dJ((mf`IVJPah@Pg@o@hZ_<fP;MGRNb?XS8#4KWB)*F6i(
z60fBXq|-#DD!%qFyc#7`L!hOrpLUNsY{PU)B)yRV39u-0el3_1@MEX}^O27_>Y42L
z^p6!n1-@u^d9Tc;D^)LjttVW-@IboD6B{&oY~dE`SVAk!sFS(Nf}3Vn=|$GImP%U<
z*J*5Yz4zfk{=@{K6{Z>3uEE~ur?*a6w0G!;oUnWMI<CN}GuBHl0d4MC>zb6N2p@8y
z#cB}Ek|+axSC6wF&JiBtHRdaX&n>aRwr8dPPrrDtUcdNER?O=0MYP^arcCm7|K4@Z
G%Kru+)hsyx
diff --git a/predihood/main.py b/predihood/main.py
index a3556fe8..54771dc0 100644
--- a/predihood/main.py
+++ b/predihood/main.py
@@ -28,8 +28,10 @@ url = "http://127.0.0.1:8081/"
def index(page):
"""
Render the main page of the interface, i.e. `index.html`.
+
Args:
page: The name of the specific page to display.
+
Returns:
The page to display.
"""
@@ -42,6 +44,7 @@ def index(page):
def get_algorithms_page():
"""
Render the page of the algorithmic interface, i.e. `algorithms.html`.
+
Returns:
The page to display.
"""
@@ -52,6 +55,7 @@ def get_algorithms_page():
def get_details_iris():
"""
Get all information about the given IRIS.
+
Returns:
A page that contains all information about the given IRIS (descriptive, grouped and raw indicators).
"""
@@ -67,6 +71,7 @@ def get_details_iris():
def get_classifiers():
"""
Get list of available classifiers (stored in AVAILABLE_CLASSIFIERS).
+
Returns:
A list containing the names of the available classifiers.
"""
@@ -75,6 +80,12 @@ def get_classifiers():
@app.route('/getParameters', methods=["GET"])
def get_parameters():
+ """
+ Get parameters of the given classifier by give its name.
+
+ Returns:
+ A dictionary containing for each parameter its name, its types, its default value, and its definition.
+ """
if 'name' in request.args:
name = request.args['name']
parameters = signature(name)
@@ -88,6 +99,7 @@ def get_parameters():
def run_algorithm():
"""
Run classifier on data with the specified parameters.
+
Returns:
The computed accuracies for each EV and each list. The top-k are also returned.
"""
@@ -115,6 +127,7 @@ def run_algorithm():
def predict_iris():
"""
Predict the environment (i.e. 6 EV) of the given IRIS.
+
Returns:
Predictions for each EV
"""
@@ -130,6 +143,12 @@ def predict_iris():
@app.route('/getIrisPolygon', methods=["GET"])
def get_iris_for_polygon():
+ """
+ Get the list of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ A list of the IRIS that are in the given polygon.
+ """
lat1 = float(request.args['lat1'])
lng1 = float(request.args['lng1'])
lat2 = float(request.args['lat2'])
@@ -144,6 +163,12 @@ def get_iris_for_polygon():
@app.route('/countIrisPolygon', methods=["GET"])
def count_iris_for_polygon():
+ """
+ Count the number of IRIS in the given polygon. The polygon is defined by two points given in the AJAX request.
+
+ Returns:
+ The number of IRIS that are in the given polygon.
+ """
lat1 = float(request.args['lat1'])
lng1 = float(request.args['lng1'])
lat2 = float(request.args['lat2'])
@@ -154,6 +179,12 @@ def count_iris_for_polygon():
@app.route('/searchCode', methods=["GET"])
def get_iris_from_code():
+ """
+ Get an IRIS object (represented by a dictionary) given its code. The code is a string of 9 digits.
+
+ Returns:
+ An object that represents the IRIS corresponding to the given code.
+ """
code_iris = request.args['codeIris']
iris = model.get_iris_from_code(code_iris)
if iris is None:
@@ -165,6 +196,12 @@ def get_iris_from_code():
@app.route('/searchName', methods=["GET"])
def get_iris_from_name():
+ """
+ Get IRIS given its name. The search is done on IRIS's name and IRIS' city.
+
+ Returns:
+ A list of IRIS corresponding to the given name.
+ """
query = request.args['querySearch']
iris = model.get_iris_from_name(query)
if iris is None or len(iris) == 0:
@@ -174,19 +211,14 @@ def get_iris_from_name():
return json.dumps({'status': 'OK', 'geojson': iris})
-@app.route('/getEnvironmentValues', methods=["GET"])
-def get_environment_values():
- variables_with_values = {}
- for env in ENVIRONMENT_VALUES:
- temp = []
- for key in ENVIRONMENT_VALUES[env]:
- temp.append(ENVIRONMENT_VALUES[env][key]) # get english values
- variables_with_values[env] = temp
- return json.dumps(variables_with_values) # {"result": variables_with_values}
-
-
@app.route('/add_iris_to_csv', methods=["GET"])
def add_iris_to_csv():
+ """
+ Adds an assessed IRIS to a CSV file. Not available in production mode.
+
+ Returns:
+ A message that explain the status of the request, i.e. OK if the IRIS has been added, KO else.
+ """
assessed_values = []
for env in ENVIRONMENT_VALUES:
assessed_values.append(request.args[env])
@@ -197,6 +229,12 @@ def add_iris_to_csv():
@app.route('/favicon.ico')
@app.route('/<page>/favicon.ico')
def favicon():
+ """
+ Display the favicon.
+
+ Returns:
+ The favicon.
+ """
return send_from_directory(os.path.join(app.root_path, 'static'), 'favicon.png', mimetype='image/favicon.png')
diff --git a/predihood/model.py b/predihood/model.py
index 748fe7bc..24917137 100644
--- a/predihood/model.py
+++ b/predihood/model.py
@@ -19,13 +19,15 @@ json_iris_indicator_code_to_label = 'static/data/dictionnaire-indicateurs.json'
def get_iris_for_polygon(lat1, lng1, lat2, lng2):
"""
Get IRIS that are in a box represented by given coordinates.
+
Args:
- lat1: A float for latitude of the first point of the box.
- lng1: A float for longitude of the first point of the box.
- lat2: A float for latitude of the second point of the box.
- lng2: A float for longitude of the second point of the box.
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
Returns:
- A list of IRIS that are in the box defined by given latitudes and longitudes.
+ a list of IRIS that are in the box defined by given latitudes and longitudes
"""
# polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
polygon = Mongiris.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
@@ -37,13 +39,15 @@ def get_iris_for_polygon(lat1, lng1, lat2, lng2):
def count_iris_for_polygon(lat1, lng1, lat2, lng2):
"""
Count IRIS that are in a box represented by given coordinates.
+
Args:
- lat1: A float for latitude of the first point of the box.
- lng1: A float for longitude of the first point of the box.
- lat2: A float for latitude of the second point of the box.
- lng2: A float for longitude of the second point of the box.
+ lat1: a float for latitude of the first point of the box
+ lng1: a float for longitude of the first point of the box
+ lat2: a float for latitude of the second point of the box
+ lng2: a float for longitude of the second point of the box
+
Returns:
- An integer representing the number of IRIS that are in the box defined by given latitudes and longitudes.
+ an integer representing the number of IRIS that are in the box defined by given latitudes and longitudes
"""
polygon = db.convert_geojson_box_to_polygon(lng1, lat1, lng2, lat2)
iris = db.geo_within(iris_collection, polygon)
@@ -54,10 +58,12 @@ def count_iris_for_polygon(lat1, lng1, lat2, lng2):
def get_iris_from_code(code_iris):
"""
Get an IRIS given its code (9 digits).
+
Args:
- code_iris: A string corresponding to the code of the IRIS. The code should be a 9 digits string.
+ code_iris: a string corresponding to the code of the IRIS (the code should be a 9 digits string)
+
Returns:
- An object that represents the IRIS corresponding to the given code.
+ an object that represents the IRIS corresponding to the given code
"""
iris = db.get_iris_from_code(code_iris)
return iris
@@ -66,10 +72,12 @@ def get_iris_from_code(code_iris):
def get_iris_from_name(name):
"""
Get an IRIS given its name.
+
Args:
- name: A string corresponding to the name of the IRIS.
+ name: a string corresponding to the name of the IRIS
+
Returns:
- An object that represents the IRIS corresponding to the given name.
+ an object that represents the IRIS corresponding to the given name
"""
# the query string (name) is searched in both the name of the iris and the name of the city
regx = re.compile(name, re.IGNORECASE)
@@ -81,10 +89,12 @@ def get_iris_from_name(name):
def parse_json_to_dict(json_file_path):
"""
Convert a JSON file to a dictionary object.
+
Args:
- json_file_path: A string containing the path of the file to load.
+ json_file_path: a string containing the path of the file to load
+
Returns:
- A dictionary containing the data in the file located in the given path.
+ a dictionary containing the data in the file located in the given path
"""
assert(os.path.isfile(json_file_path))
with open(json_file_path) as data_file:
@@ -96,10 +106,12 @@ def parse_json_to_dict(json_file_path):
def get_coords_from_code(code):
"""
Get geometry of the IRIS corresponding to the given code.
+
Args:
- code: A string corresponding to the code of the IRIS. The code should be a 9 digits string.
+ code: a string corresponding to the code of the IRIS. The code should be a 9 digits string
+
Returns:
- A list representing coordinates of the IRIS.
+ a list representing coordinates of the IRIS
"""
iris = db.get_iris_from_code(code)
if iris:
@@ -111,8 +123,9 @@ def get_coords_from_code(code):
def get_indicators_list():
"""
Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
Returns:
- A list containing all names of indicators, e.g. ['POP0002', 'POP0204', ..., 'P14_RP_MAISON'].
+ a list containing all names of indicators, e.g. ['POP0002', 'POP0204', ..., 'P14_RP_MAISON']
"""
list_indicators = db.find_all(db.collection_indic)
return [indicator["short_label"] for indicator in list_indicators]
@@ -121,8 +134,9 @@ def get_indicators_list():
def get_indicators_dict():
"""
Get all INSEE indicators that are stored in the database. This corresponds to the collection 'collindic'.
+
Returns:
- A dictionary containing all names of indicators, e.g. {'POP0002': 'Population aged from 0 to 2 y.o.', ..., 'P14_RP_MAISON': 'Number of principal residences'}.
+ a dictionary containing all names of indicators, e.g. {'POP0002': 'Population aged from 0 to 2 y.o.', ..., 'P14_RP_MAISON': 'Number of principal residences'}
"""
list_indicators = db.find_all(db.collection_indic)
return {indicator["short_label"]: indicator["full_label"] for indicator in list_indicators}
diff --git a/predihood/predict.py b/predihood/predict.py
index 7f85ce08..d67729b2 100644
--- a/predihood/predict.py
+++ b/predihood/predict.py
@@ -4,7 +4,7 @@ from collections import OrderedDict
from predihood.classes.Data import Data
from predihood.classes.Dataset import Dataset
from predihood.classes.MethodPrediction import MethodPrediction
-from predihood.config import ENVIRONMENT_VARIABLES, TOPS_K, FOLDER_SELECTED_INDICATORS, RANDOM_STATE, FILE_LIST_DISTRIBUTION
+from predihood.config import ENVIRONMENT_VARIABLES, TRAIN_SIZE, TEST_SIZE
from predihood.selection import retrieve_lists
from predihood.utility_functions import check_dataset_size, get_most_frequent
from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
@@ -33,15 +33,17 @@ CLASSIFIERS = [
def compute_all_accuracies(data, clf, train_size, test_size, remove_outliers=False, remove_rural=False):
"""
Compute accuracies for each EV and each list with the given classifier.
+
Args:
- data: a Data object that contains assessed IRIS and their attributes.
- clf: an object which is a classifier (with `fit` and `predict` methods).
- train_size: the size of the training sample.
- test_size: the size of the test sample.
+ data: a Data object that contains assessed IRIS and their attributes
+ clf: an object which is a classifier (with `fit` and `predict` methods)
+ train_size: an integer corresponding to the size of the training sample
+ test_size: an integer corresponding to the size of the test sample
remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
- remove_rural: True to remove IRIS that are in the countryside to avoid bias while predicting
+ remove_rural: True to remove IRIS that are in the countryside to avoid bias while predicting, False else
+
Returns:
- A dict of results for each EV and each list of selected indicators.
+ a dictionary of results for each EV and each list of selected indicators
"""
log.info("... Computing accuracies ...")
train_size, test_size = check_dataset_size(train_size, test_size)
@@ -86,18 +88,20 @@ def compute_all_accuracies(data, clf, train_size, test_size, remove_outliers=Fal
return results
-def predict_one_iris(iris_code, data, clf, train_size, test_size, remove_outliers):
+def predict_one_iris(iris_code, data, clf, train_size, test_size, remove_outliers=False):
"""
Predict the 6 EV for the given IRIS, the given data and the given classifier.
+
Args:
- iris_code: A string that contains the code of the IRIS (9 digits).
- data: A Data object on which classifier will learn.
- clf: An object (classifier) to perform the prediction.
- train_size: The size of the train sample.
- test_size: The size of the test sample.
- remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else.
+ iris_code: a string that contains the code of the IRIS (9 digits)
+ data: a Data object on which classifier will learn
+ clf: an object (classifier) to perform the prediction
+ train_size: an integer corresponding to the size of the train sample
+ test_size: an integer corresponding to the size of the test sample
+ remove_outliers: True to remove from the dataset IRIS that are detected as outliers, False else
+
Returns:
- Predictions for each EV.
+ A dictionary containing predictions for each EV.
"""
train_size, test_size = check_dataset_size(train_size, test_size)
lists = retrieve_lists()
@@ -109,40 +113,23 @@ def predict_one_iris(iris_code, data, clf, train_size, test_size, remove_outlier
for top_k, lst in lists.items():
dataset = Dataset(data, env, selected_indicators=lst[env], train_size=train_size, test_size=test_size, outliers=remove_outliers, _type="supervised")
dataset.init_all_in_one()
- algo = MethodPrediction(name='', dataset=dataset, classifier=clf)
- algo.fit()
- algo.predict(iris_code)
- predictions_lst.append(algo.prediction)
+ algorithm = MethodPrediction(name='', dataset=dataset, classifier=clf)
+ algorithm.fit()
+ algorithm.predict(iris_code)
+ predictions_lst.append(algorithm.prediction)
predictions[env] = get_most_frequent(predictions_lst) # get the most frequent value and the number of occurrences
- print(predictions) # TODO
+ print(predictions) # TODO: give an example of the dictionary
return predictions
if __name__ == '__main__':
- # data = Data()
- # data.init_all_in_one()
- # dataset = Dataset(data, "geo")
- # dataset.get_environment_variable()
# Create data
- # data = Data(normalization=None, filtering=False)
- # data.init_all_in_one()
- #
- # data = Data(normalization="density", filtering=False)
- # data.init_all_in_one()
- #
- # data = Data(normalization="population", filtering=False)
- # data.init_all_in_one()
- #
- # data = Data(normalization=None, filtering=True)
- # data.init_all_in_one()
-
data = Data(normalization="population", filtering=True)
data.init_all_in_one()
- # # Run expes on data
- # expe1(data)
- # expe2(data)
- # expe3(data)
- compute_all_accuracies(data, RandomForestClassifier(), 0.8, 0.2)
- # expe5(data, RandomForestClassifier(), 0.8, 0.2)
+ # Compute accuracies for each EV and each top-k
+ compute_all_accuracies(data, RandomForestClassifier(), TRAIN_SIZE, TEST_SIZE)
+
+ # Predict EV of the "Part-Dieu" IRIS, which is he CBD of Lyon (Central Business District)
+ predict_one_iris("693830301", data, RandomForestClassifier(), TRAIN_SIZE, TEST_SIZE)
diff --git a/predihood/selection.py b/predihood/selection.py
index dbfe22e6..3dc25892 100644
--- a/predihood/selection.py
+++ b/predihood/selection.py
@@ -7,26 +7,17 @@ import pandas as pd
from predihood.classes.Data import Data
from predihood.classes.Dataset import Dataset
from predihood.classes.MethodSelection import MethodSelection
-from predihood.config import TOPS_K, ENVIRONMENT_VARIABLES, FOLDER_SELECTED_INDICATORS, FILE_HIERARCHY, FILE_LIST_DISTRIBUTION
-from predihood.utility_functions import apply_hierarchy, union
+from predihood.config import TOPS_K, ENVIRONMENT_VARIABLES, FOLDER_SELECTED_INDICATORS, FILE_HIERARCHY
+from predihood.utility_functions import apply_hierarchy
from sklearn.ensemble import RandomForestClassifier, ExtraTreesClassifier
log = logging.getLogger(__name__)
-DEFAULTS = {
- "threshold_HM": 0.4,
- "min_col_HM": 0.5
-}
-
-PARAMETERS = {
- "threshold_HM": DEFAULTS["threshold_HM"],
- "min_col_HM": DEFAULTS["min_col_HM"]
-}
-
def generate_all_data():
"""
- Generate all datasets, i.e. one with density normalization, one for population normalization and one without normalization. The three are filtered.
+ Generate all datasets, i.e. one with density normalization, one for population normalization and one without normalization.
+ The three are filtered. Generated datasets are located in generated_files/datasets.
"""
data = Data(normalization="density", filtering=True)
data.init_all_in_one()
@@ -38,75 +29,97 @@ def generate_all_data():
def retrieve_lists(top_k=None):
"""
- Get all lists generated for each top-k.
- :param top_k: Specify a top_k to get the list of the top_k.
- :return: a dict containing the lists of indicators for each top-k
+ Get all lists generated for each top-k or for the given top-k if provided.
+
+ Args:
+ top_k: a integer corresponding to the top-k of the list to retrieve
+
+ Returns:
+ a dictionary containing the lists of indicators for each top-k
"""
lists = {}
if top_k:
- if not os.path.exists(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv")): generate_lists()
- indicators_csv = pd.read_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv"), header=None)
- list = json.loads(indicators_csv.drop(indicators_csv.columns[0], axis=1).to_json(orient="index")) # drop([0]) to remove column with row indexes, axis=1 to delete column
- list_temp = {}
- for key, value in list.items():
- list_temp[ENVIRONMENT_VARIABLES[int(key)]] = [value2 for key2, value2 in value.items() if value2 is not None]
- lists[str(top_k)] = list_temp
+ # a top-k is specified, so the list of indicators of size top-k is retrieved
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
else:
+ # no top-k is provided, so all lists form 10 to 100 indicators are retrieved
for top_k in TOPS_K:
- if not os.path.exists(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv")): generate_lists()
- indicators_csv = pd.read_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv"), header=None)
- list = json.loads(indicators_csv.drop(indicators_csv.columns[0], axis=1).to_json(orient="index")) # drop([0]) to remove column with row indexes, axis=1 to delete column
- list_temp = {}
- for key, value in list.items():
- list_temp[ENVIRONMENT_VARIABLES[int(key)]] = [value2 for key2, value2 in value.items() if value2 is not None]
- lists[str(top_k)] = list_temp
+ lst = retrieve_one_list(top_k)
+ lists[str(top_k)] = lst
return lists
+def retrieve_one_list(top_k):
+ """
+ Retrieve the list of selected INSEE indicators corresponding to the given top-k.
+
+ Args:
+ top_k: an integer corresponding to the size of the list to retrieve
+
+ Returns:
+ a list containing indicators of the list of size top-k
+ """
+ if not os.path.exists(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv")): generate_lists()
+ indicators_csv = pd.read_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list" + str(top_k) + ".csv"), header=None)
+ lst = json.loads(indicators_csv.drop(indicators_csv.columns[0], axis=1).to_json(orient="index"))
+ list_temp = {}
+ for key, value in lst.items():
+ list_temp[ENVIRONMENT_VARIABLES[int(key)]] = [value2 for key2, value2 in value.items() if value2 is not None]
+ return list_temp
+
+
def generate_lists():
+ """
+ Generates lists of INSEE indicators that are relevant for prediction.
+ This selection process is based on:
+ - removing fully correlated indicators
+ - select a limited number among the most relevant indicators (using Random Forest and Extra Tree classifiers)
+ - taking into account the diversity of categories of INSEE indicators based on a hierarchy of these indicators
+ """
# 1. Create data
data = Data(normalization="density", filtering=True)
data.init_all_in_one()
- # # 2. Run heat map and get less correlated indicators
- dataset = Dataset(data, "batiment", "unsupervised")
+ # 2. Run heat map and get fully correlated indicators by using a correlation matrtix
+ dataset = Dataset(data, "building_type", "unsupervised")
dataset.init_all_in_one()
- heat_map = MethodSelection(name="heat map EV-agnostic", dataset=dataset, parameters=PARAMETERS)
- heat_map.results()
- # heat_map.draw_and_save()
+ heat_map = MethodSelection(name="heat map", dataset=dataset, parameters={"method": "spearman"})
+ heat_map.compute_selection()
fully_correlated_indicators = heat_map.best_indicators
- log.info("fully correlated indicators: %d %s", len(fully_correlated_indicators), fully_correlated_indicators)
+ log.info("fully correlated indicators: %d %s", len(fully_correlated_indicators), ", ".join(fully_correlated_indicators))
+ # 3. Select a limited number (top-k) of indicators by using Random Forest and Extra Tree classifiers.
+ # Then take into account the diversity of indicators by using a hierarchy of the INSEE indicators
hierarchy = pd.read_csv(FILE_HIERARCHY, sep="\t")
- # # 4. Run Feature selection and heat map EV-oriented for each EV
- # # to construct primary and secondary lists of selected indicators # TODO
+ # 4. For each size of list (i.e. for each top-k), select a limited number of indicators with Extra Tree and Random Forest classifiers
+ # Then merge the two results to obtain a single list of relevant indicators.
for top_k in TOPS_K:
log.info("constructing list of %d indicators", top_k)
all_lists = [] # to keep lists of indicators for each EV (for the current top-k)
- PARAMETERS["top_k"] = top_k
for env in ENVIRONMENT_VARIABLES:
- # B. FEATURE IMPORTANCE on uncorrelated indicators (the ones that are not chosen by heat map) to select the most relevant ones
- dataset = Dataset(data, env, indicators_to_remove=fully_correlated_indicators, _type="supervised") # WARNING: fill _type parameter
+ dataset = Dataset(data, env, indicators_to_remove=fully_correlated_indicators, _type="supervised")
dataset.init_all_in_one()
- # a. get best indicators for ET
- fi_et = MethodSelection(name="feature importance ET", dataset=dataset, classifier=ExtraTreesClassifier(), parameters=PARAMETERS)
+ # a. get best indicators according to Extra Tree classifier
+ fi_et = MethodSelection(name="feature importance ET", dataset=dataset, classifier=ExtraTreesClassifier(), parameters={"top_k": top_k})
fi_et.fit()
- fi_et.results()
- best_indicators_FI_ET = fi_et.best_indicators
+ fi_et.compute_selection()
+ best_indicators_ET = fi_et.best_indicators
- # b. get best indicators for RF
- fi_rf = MethodSelection(name="feature importance RF", dataset=dataset, classifier=RandomForestClassifier(), parameters=PARAMETERS)
+ # b. get best indicators according to Random Forest classifier
+ fi_rf = MethodSelection(name="feature importance RF", dataset=dataset, classifier=RandomForestClassifier(), parameters={"top_k": top_k})
fi_rf.fit()
- fi_rf.results()
- best_indicators_FI_RF = fi_rf.best_indicators
+ fi_rf.compute_selection()
+ best_indicators_RF = fi_rf.best_indicators
- # c. merge best indicators from ET and RF
- best_indicators_FI_ET.extend(best_indicators_FI_RF)
- all_selected_indicators = best_indicators_FI_ET # all selected indicators, i.e. union between RF and ET
+ # c. merge indicators that have been selected by ET and RF classifiers
+ # in this step, if an indicator have been selected by the two classifiers, its score is the addition of its score for RF and the one for ET.
+ best_indicators_ET.extend(best_indicators_RF)
+ all_selected_indicators = best_indicators_ET # all selected indicators, i.e. union between RF and ET
keys = set([element[0] for element in all_selected_indicators]) # store indicators' names selected
merged_indicators_temp = {key: 0 for key in keys}
@@ -114,22 +127,22 @@ def generate_lists():
indicator = all_selected_indicators[i]
merged_indicators_temp[indicator[0]] += indicator[1] # adding score
- # transform it into a list of sublists, e.g. [[indicator1, score1], ..., indicatorN, scoreN]]
+ # transform it into a list of sub-lists, e.g. [[indicator1, score1], ..., indicatorN, scoreN]]
merged_indicators = [[key, merged_indicators_temp[key]] for key in merged_indicators_temp]
- # d. apply hierarchy on selected indicators to keep the best ancestors
- selected_indicators = apply_hierarchy(merged_indicators, hierarchy)
+ # d. apply hierarchy on selected indicators to taking into account the diversity of categories of indicators
+ indicators_hierarchy = apply_hierarchy(merged_indicators, hierarchy)
# get the names of each selected indicator
- selected_indicators_names = [indic[0] for indic in selected_indicators]
+ selected_indicators_names = [indicator[0] for indicator in indicators_hierarchy]
# e. add uncorrelated indicators of heat map to the lists
all_lists.append(selected_indicators_names)
# C. Transform lists of selected indicators for the current top-k and save it as a CSV file
# indexes = {i: ENVIRONMENT_VARIABLES[i] for i in range(len(ENVIRONMENT_VARIABLES))}
- selected_indicators_prim = pd.DataFrame(np.array(all_lists).tolist(), index=None, columns=None)
- selected_indicators_prim.to_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list"+str(top_k)+".csv"), header=False)
+ selected_indicators = pd.DataFrame(np.array(all_lists).tolist())
+ selected_indicators.to_csv(os.path.join(FOLDER_SELECTED_INDICATORS, "list"+str(top_k)+".csv"), header=False)
if __name__ == "__main__":
diff --git a/predihood/utility_functions.py b/predihood/utility_functions.py
index d282e679..178055e5 100644
--- a/predihood/utility_functions.py
+++ b/predihood/utility_functions.py
@@ -2,10 +2,7 @@ import ast
import inspect
import json
import logging
-import math
-import matplotlib.pyplot as plt
import numpy as np
-import os
import pandas as pd
import re
import requests
@@ -13,58 +10,77 @@ import stringdist
from area import area
from predihood import model
+from predihood.config import AVAILABLE_CLASSIFIERS, TRAIN_SIZE, TEST_SIZE, ENVIRONMENT_VARIABLES, FOLDER_DATASETS, FOLDER_DISTRIBUTION, TRANSLATION, OLD_PREFIX, NEW_PREFIX, FILE_MANUAL_ASSESSMENT
log = logging.getLogger(__name__)
-#################### iris function ####################
-from predihood.config import AVAILABLE_CLASSIFIERS, TRAIN_SIZE, TEST_SIZE, ENVIRONMENT_VARIABLES, FOLDER_DATASETS, FOLDER_DISTRIBUTION, TRANSLATION, OLD_PREFIX, NEW_PREFIX, FILE_MANUAL_ASSESSMENT
-
-
+# 1. IRIS function
def address_to_code(address):
"""
Get IRIS code from address.
- :param address: the address of the iris.
- :return: the code of the iris.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the code of the searched address
"""
- response = requests.get("https://pyris.datajazz.io/api/search/", params=[('q', address)])
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
json_response = response.json()
return str(json_response["complete_code"]) if "complete_code" in json_response else None
def address_to_city(address):
- response = requests.get("https://pyris.datajazz.io/api/search/", params=[('q', address)])
+ """
+ Get city from address.
+
+ Args:
+ address: a string containing the address of the iris
+
+ Returns:
+ a string containing the city of the searched address
+ """
+ response = requests.get("https://pyris.datajazz.io/api/search/", params=[("q", address)])
json_response = response.json()
return str(json_response["name"]) if "name" in json_response else None
-def append_indicator(raw_indicator, iris, arr, append_col, indicators):
+def append_indicator(raw_indicator, iris, lst, append_col, indicators):
"""
- Append an indicator in the current array (departure or arrival).
- :param raw_indicator: the name of the indicator, e.g. P14_POP.
- :param iris: the iris object that contains the indicator.
- :param arr: the array where to append the indicator's value.
- :param append_col: True to append the indicator to the cols variable.
- :param indicators: the list of the columns of the dataset, i.e. the indicators.
- :return: the array containing indicator' values and the other one containing indicators names.
+ Append the value of an INSEE indicator to a list corresponding to the data of the assessed IRIS (departure or arrival).
+
+ Args:
+ raw_indicator: a string containing the name of the indicator, e.g. P14_POP
+ iris: the iris object that contains the indicator
+ lst: the list where to append the indicator's value
+ append_col: True to append the indicator to the cols variable
+ indicators: the list of the columns of the dataset, i.e. the indicators
+
+ Returns:
+ the list containing indicator' values and the other one containing indicators names.
"""
if raw_indicator in iris["properties"]["raw_indicators"]:
val_indicator = iris["properties"]["raw_indicators"][raw_indicator]
- arr.append(np.nan) if (val_indicator is None or not val_indicator) else arr.append(
- val_indicator) # append 0 if indicator is null, else append the value
+ # append NaN if indicator is null, else append the value
+ lst.append(np.nan) if (val_indicator is None or not val_indicator) else lst.append(val_indicator)
else:
- arr.append(np.nan)
+ lst.append(np.nan)
if append_col: indicators.append(raw_indicator)
- return arr, indicators
+ return lst, indicators
-def append_target(row, target, arr):
+def append_target(row, target, lst):
"""
- Append the target to the current array (departure or arrival).
- :param row: the current row of dataset.
- :param target: the target to append, i.e. the value of the environment variable.
- :param arr: the array where the target is append.
- :return: the array with the appended target.
+ Append the target to the current list (departure or arrival).
+
+ Args:
+ row: a list corresponding to the current row of dataset
+ target: a string containing the target to append, i.e. the value of the EV
+ lst: the list where the target is append
+
+ Returns:
+ the list with the appended target
"""
if target == OLD_PREFIX + "geographical_position" or target == NEW_PREFIX + "geographical_position":
# get only geo position without the city, e.g. South-East Lyon gives South-East
@@ -73,20 +89,24 @@ def append_target(row, target, arr):
else:
city = row[target].split(" ")[0]
# city_name = ''.join(split_city[1:len(split_city)])
- value = TRANSLATION[city] if city in TRANSLATION else np.nan # + " " + city_name
- arr.append(value) # if (value is None or not value or str(value) == "nan") else arr.append(value)
+ value = TRANSLATION[city] if city in TRANSLATION else np.nan
+ lst.append(value)
else:
value = TRANSLATION[row[target]] if row[target] in TRANSLATION else np.nan
- arr.append(value) # if (value is None or not value or str(value) == "nan") else arr.append(value)
+ lst.append(value)
- return arr
+ return lst
def indicator_full_to_short_label(full_label):
"""
Convert the full label of an indicator to its short label.
- :param full_label: the full label of the indicator, e.g. Population 0-2 y.o. in 2014.
- :return: the short label of the indicator, e.g. P14_POP0002.
+
+ Args:
+ full_label: a string containing the full label of the indicator, e.g. Population 0-2 y.o. in 2014
+
+ Returns:
+ the short label of the given indicator, e.g. P14_POP0002
"""
indicators = model.get_indicators_dict()
key_list = list(indicators.keys())
@@ -100,24 +120,32 @@ def indicator_full_to_short_label(full_label):
def indicator_short_to_full_label(short_label):
"""
Convert the short label of an indicator to its full label.
- :param short_label: the short label of the indicator, e.g. P14_POP0002.
- :return: the full label of the indicator, e.g. Population 0-2 y.o. in 2014.
+
+ Args:
+ short_label: a string containing the short label of the indicator, e.g. P14_POP0002
+
+ Returns:
+ the full label of the indicator, e.g. Population 0-2 y.o. in 2014
"""
indicators = model.get_indicators_dict()
return indicators[short_label]
-def apply_hierarchy(best_indicators, hierarchy):
+def apply_hierarchy(selected_indicators, hierarchy):
"""
Apply hierarchy on the given indicators in order to go up children indicators in their parents.
- :param best_indicators: the best indicators selected by the feature importance.
- :param hierarchy: the hierarchy of the indicators, for each indicator, there are its level in the hierarchy and its first ancestor.
- :return: the new best indicators list, with some children go up in their parents.
+
+ Args:
+ selected_indicators: a list of indicators selected by the feature importance process.
+ hierarchy: the hierarchy of the indicators, i.e. for each indicator, there are its level in the hierarchy and its first ancestor.
+
+ Returns:
+ the new selected indicators list, with some children which have go up in their parents.
"""
- list_indicators_FI = [best_indicators[j][0] for j in range(len(best_indicators))]
+ list_indicators_FI = [selected_indicators[j][0] for j in range(len(selected_indicators))]
indexes_to_remove = []
- for i in range(len(best_indicators)):
- index_row = hierarchy.index[hierarchy['INDICATOR'] == best_indicators[i][0]].tolist()
+ for i in range(len(selected_indicators)):
+ index_row = hierarchy.index[hierarchy["INDICATOR"] == selected_indicators[i][0]].tolist()
if len(index_row) == 0:
continue # pass this indicator because it does not exist in the hierarchy
else:
@@ -128,66 +156,79 @@ def apply_hierarchy(best_indicators, hierarchy):
if ancestor in list_indicators_FI:
index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
while index_ancestor in indexes_to_remove:
- ancestor2 = hierarchy.iloc[hierarchy.index[hierarchy['INDICATOR'] == ancestor].tolist()[0], 3] # name of ancestor
+ ancestor2 = hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 3] # name of ancestor
ancestor = ancestor2
index_ancestor = list_indicators_FI.index(ancestor) if ancestor in list_indicators_FI else None
- if hierarchy.iloc[hierarchy.index[hierarchy['INDICATOR'] == ancestor].tolist()[0], 2] == 1:
+ if hierarchy.iloc[hierarchy.index[hierarchy["INDICATOR"] == ancestor].tolist()[0], 2] == 1:
break
if index_ancestor not in indexes_to_remove:
- best_indicators[index_ancestor][1] += best_indicators[i][1]
+ selected_indicators[index_ancestor][1] += selected_indicators[i][1]
indexes_to_remove.append(i)
- for index in sorted(indexes_to_remove,
- reverse=True): # remove in reverse order to do not throw off subsequent indexes
- del best_indicators[index]
- return best_indicators
+ for index in sorted(indexes_to_remove,reverse=True): # remove in reverse order to do not throw off subsequent indexes
+ del selected_indicators[index]
+ return selected_indicators
def get_classifier(name):
"""
Get an instance of a classifier with its name.
- :param name: a string containing the name of the desired classifier
- :return: an instance of a classifier
+
+ Args:
+ name: a string containing the name of the desired classifier
+
+ Returns:
+ an instance of a classifier
"""
classifier = AVAILABLE_CLASSIFIERS[name]
return classifier()
-def set_classifier(clf, parameters):
+def set_classifier(classifier, parameters):
"""
- Tune the classifier with the user's parameters.
- :param clf: instance of classifier to tune
- :param parameters: a dict containing the tuning parameters
- :return: an instance of the tuned classifier
+ Tune the classifier with the given parameters.
+
+ Args:
+ classifier: an instance of the classifier to tune
+ parameters: a dictionary containing the tuning parameters
+
+ Returns:
+ an instance of the tuned classifier
"""
- keys_clf = list(clf.get_params().keys())
+ keys_clf = list(classifier.get_params().keys())
# remove None parameters and parameters that don't exist in sklearn (e.g. train and test size)
parameters = {key: value for key, value in parameters.items() if value != "" and key in keys_clf}
- clf.set_params(**parameters)
- return clf
+ classifier.set_params(**parameters)
+ return classifier
def signature(chosen_algorithm):
"""
- Get the signature of an algorithm, i.e. its parameters, the default values and the type of each parameter.
- :param chosen_algorithm: the name of the algorithm in str, e.g. RandomForestClassifier
- :return: the instance of the classifier and a dict containing each parameter associated with its default value and the type of the parameter.
+ Get the signature of an algorithm, i.e. its parameters, the default values and the type of each parameter. The documentation of the algorithm must be in NumPy style.
+
+ Args:
+ chosen_algorithm: the name of the algorithm in str, e.g. RandomForestClassifier
+
+ Returns:
+ the signature of the given algorithm, i.e. a dictionary containing for each parameter:
+ - a list of the accepted types
+ - the default value
+ - a description of the parameter (e.g. "The train_size parameter aims at tuning the size of the sample during the learning step.")
"""
# special case for no selection
if chosen_algorithm == "Algorithm": return json.dumps({})
try:
# model = eval(_chosen_algorithm) # never use eval on untrusted strings
model = get_classifier(chosen_algorithm)
- doc = model.__doc__
+ doc = model.__doc__ # TODO: specify case when there is no doc (user-implemented algorithm)
param_section = "Parameters"
dashes = "-" * len(param_section) # -------
number_spaces = doc.find(dashes) - (doc.find(param_section) + len(param_section))
attribute_section = "Attributes\n"
# sub_doc is the param section of the docs (i.e. without attributes and some text)
- sub_doc = doc[doc.find(param_section) + len(param_section) + number_spaces + len(dashes) + len("\n"):doc.find(
- attribute_section)]
+ sub_doc = doc[doc.find(param_section) + len(param_section) + number_spaces + len(dashes) + len("\n"):doc.find(attribute_section)]
except:
raise Exception("This algorithm does not exist for the moment...")
- params = inspect.getfullargspec(model.__init__).args[1:] # get parameter' names -- [1:] to remove 'self' parameter
+ params = inspect.getfullargspec(model.__init__).args[1:] # get parameter' names -- [1:] to remove self parameter
defaults = inspect.getfullargspec(model.__init__).defaults # get default values
assert len(params) == len(defaults)
parameters = {}
@@ -197,36 +238,36 @@ def signature(chosen_algorithm):
index_next_newline = sub_doc[index_param:].find("\n") # find returns the first occurrence
parameter_string = sub_doc[index_param:index_param + index_next_newline]
doc_param = sub_doc[index_param + index_next_newline:]
- index_end_sentence = re.search('(\.\s)', doc_param).start() # search for the first sentence
+ index_end_sentence = re.search("(\.\s)", doc_param).start() # search for the first sentence
first_sentence = doc_param[:index_end_sentence + 1]
# format first sentence to have a prettier display.
first_sentence = first_sentence.replace("\n", " ")
while " " in first_sentence:
first_sentence = first_sentence.replace(" ", " ")
types_and_default = parameter_string[len(param_name):]
- if '{' in types_and_default and '}' in types_and_default: # for cases like {'auto', 'kd_tree', 'brute'}, optional
- types_and_default = types_and_default.replace('{', '')
- types_and_default = types_and_default.replace('}', '')
- if ' or ' in types_and_default: types_and_default = types_and_default.replace(' or ', ', ')
- types_defaults_split = types_and_default.split(', ')
+ if "{" in types_and_default and "}" in types_and_default: # for cases like {"auto", "kd_tree", "brute"}, optional
+ types_and_default = types_and_default.replace("{", '')
+ types_and_default = types_and_default.replace("}", '')
+ if " or " in types_and_default: types_and_default = types_and_default.replace(" or ", ", ")
+ types_defaults_split = types_and_default.split(", ")
types = []
default = -1
- variants = ['optional (default=', 'optional (default = ', 'optional', '(default=', '(default = ', 'default ',
- 'default: ', 'default='] # DO NOT CHANGE THE ORDER OF ITEMS
+ variants = ["optional (default=", "optional (default = ", "optional", "(default=", "(default = ", "default ",
+ "default: ", "default="] # DO NOT CHANGE THE ORDER OF ITEMS
for item in types_defaults_split:
if not any(value in item for value in variants):
- if item.startswith('length'):
+ if item.startswith("length"):
pass # exceptions
else:
types.append(item) # item is a type
else:
for value in variants:
if value in item:
- if value.startswith('optional ('):
+ if value.startswith("optional ("):
default = item.split(value)[1][:-1]
- elif value.startswith('(default'):
+ elif value.startswith("(default"):
default = item.split(value)[1][:-1]
- elif value.startswith('default'):
+ elif value.startswith("default"):
default = item.split(value)[1]
elif value == "optional":
default = "None"
@@ -235,23 +276,25 @@ def signature(chosen_algorithm):
type_of_default = str(type(ast.literal_eval(str(default))).__name__)
else:
type_of_default = "str"
- types[:] = ['int' if x == 'integer' else x for x in types] # replace 'integer' by 'int'
- types[:] = ['bool' if x == 'boolean' else x for x in types] # replace 'boolean' by 'bool'
- types[:] = ['str' if x == 'string' else x for x in types] # replace 'boolean' by 'bool'
+ types[:] = ["int" if x == "integer" else x for x in types] # replace "integer" by "int"
+ types[:] = ["bool" if x == "boolean" else x for x in types] # replace "boolean" by "bool"
+ types[:] = ["str" if x == "string" else x for x in types] # replace "str" by "string"
if len(types) == 0: types.append(type_of_default) # fill missing types
- types[:] = [x for x in types if 'None' not in x and 'NoneType' not in x] # remove None type
- description = ""
- parameters[param_name[:-3]] = {"types": types, "default": default,
- "description": first_sentence} # -3 to remove ' : '
+ types[:] = [x for x in types if "None" not in x and "NoneType" not in x] # remove None type
+ parameters[param_name[:-3]] = {"types": types, "default": default, "description": first_sentence} # -3 to remove " : "
return parameters
def check_dataset_size(train_size, test_size):
"""
Check train and test size and update with defaults or divided by 100 if needed.
- :param train_size: The value for train size (should be between 0 and 1).
- :param test_size: The value for test size (should be between 0 and 1).
- :return: The values for train and test sizes.
+
+ Args:
+ train_size: an integer or a float corresponding to the value for train size (should be between 0 and 1)
+ test_size: an integer or a float corresponding to the value for test size (should be between 0 and 1)
+
+ Returns:
+ the train and test sizes
"""
if 0 < train_size < 1 and 0 < test_size < 1 and train_size + test_size == 1:
return train_size, test_size # default case
@@ -267,15 +310,17 @@ def check_dataset_size(train_size, test_size):
return train_size, test_size
-#################### list functions ####################
-
-
+# 2. list functions
def intersection(lst1, lst2):
"""
Intersect two lists.
- :param lst1: the first list to be intersected.
- :param lst2: the second list to be intersected.
- :return: the result of the intersection of the two lists.
+
+ Args:
+ lst1: a list corresponding to the first list to be intersected
+ lst2: a list corresponding to the second list to be intersected
+
+ Returns:
+ a list corresponding to the result of the intersection of the two given lists.
"""
return list(set(lst1) & set(lst2))
@@ -283,123 +328,79 @@ def intersection(lst1, lst2):
def union(lst1, lst2):
"""
Unify two lists without repetitions.
- :param lst1: the first list to append.
- :param lst2: the second list to append.
- :return: the union of the two lists.
+
+ Args:
+ lst1: a list corresponding to the first list to append
+ lst2: a list corresponding to the second list to append
+
+ Returns:
+ a list corresponding to the union of the two lists
"""
return list(set(lst1) | set(lst2))
-def sim(val, arr):
- """Check if a value is enough similar to the data.
- :param val: the value on which the similarity is computed.
- :param arr: the array containing other values to check similarity.
- :return: the index of the similar value, -1 if no value os enough similar.
+def similarity(value, lst):
"""
- for i in range(len(arr)):
- if val in arr[i]: return -2 # this value is already append, to append it a second time
- for elem in arr[i]:
+ Check if a value is enough similar to the data.
+
+ Args:
+ value: the value on which the similarity is computed
+ lst: the list containing other values to check similarity
+
+ Returns:
+ the index of the similar value, -1 if no value is enough similar
+ """
+ for i in range(len(lst)):
+ if value in lst[i]: return -2 # this value is already append
+ for elem in lst[i]:
if not isinstance(elem, float):
- dissim = stringdist.levenshtein(str(elem), str(val))
- if dissim == 1: return i
+ dissimilarity = stringdist.levenshtein(str(elem), str(value)) # compute Levenshtein similarity
+ if dissimilarity == 1: return i # if the given value has only one difference with the current value, store it as a similarity
return -1
def get_most_frequent(lst):
- """Get the most frequent item in a list. If many elements are frequent, it returns the first one.
- :param lst: the list to find the most frequent element.
+ """
+ Get the most frequent item in a list. If many elements are frequent, it returns the first one.
+
+ Args:
+ lst: the list to find the most frequent element
+
+ Returns:
+ a dictionary containing the most frequent of the given list and its count
"""
most_frequent_element = max(set(lst), key=lst.count)
dictionary = {"most_frequent": most_frequent_element, "count_frequent": lst.count(most_frequent_element)}
return dictionary
-#################### plot functions ####################
-
-
-def auto_label(rects, ax):
+# 3. plot functions
+def auto_label(rectangles, axes):
"""
- Adds a text label above each bar in *rects*, displaying its value.
- :param rects: the bars of the plot.
- :param ax: the axes of the plot.
+ Adds a text label above each bar in rectangles, displaying its value.
+
+ Args:
+ rectangles: the bars of the plot.
+ axes: the axes of the plot.
"""
- for rect in rects:
+ for rect in rectangles:
height = rect.get_height()
- ax.annotate('{}'.format(height), xy=(rect.get_x() + rect.get_width() / 2, height), xytext=(0, 3),
- textcoords="offset points", ha='center', va='bottom')
-
-
-def draw_features_importance(importance, std, indices, X_names, env, clf, xlim_min, xlim_max, threshold):
- """
- Plot (and save) features importance in a bar chart order by importance value.
- :param importance: the importance of each value.
- :param std: the standard deviation for each feature.
- :param indices: the indices to order the plot.
- :param X_names: the names of the features.
- :param env: the name of the considered environmental variable.
- :param clf: the name of the classifier used to compute importance.
- :param threshold: the indicators under this threshold are considered not significant.
- """
- j = 0
- step = 20
- while j < len(indices):
- next_j = j + step
- if next_j >= len(indices):
- range_indices = len(indices) - j
- next_j = len(indices)
- else:
- range_indices = step
-
- # Plot the feature importance of the forest
- current_indices = np.argsort(importance)[::-1][j:next_j]
- fig, ax = plt.subplots()
- fig.set_size_inches(30, 20)
+ axes.annotate("{}".format(height), xy=(rect.get_x() + rect.get_width() / 2, height), xytext=(0, 3), textcoords="offset points", ha="center", va="bottom")
- plt.title("Importance des indicateurs INSEE selon " + str(clf))
- plt.barh(range(range_indices), importance[current_indices], yerr=std[current_indices][::-1], align="center",
- color='orange')
- plt.axvline(x=threshold, color='red', linewidth=4)
- plt.yticks(range(range_indices), [X_names[i] for i in current_indices])
- plt.ylim([-1, range_indices])
- plt.xlim([xlim_min, xlim_max])
- ax.invert_yaxis() # labels read top-to-bottom
- ax.set_xlabel('Importance')
- for i, v in enumerate(importance[current_indices]): ax.text(0, i + 0.15, str(round(v, 4)), color='black')
- if not os.path.exists(FOLDER_FEATURE_SELECTION + str(env)): os.makedirs(FOLDER_FEATURE_SELECTION + str(env))
- filename = os.path.join(FOLDER_FEATURE_SELECTION, str(env), "feature-importance-" + str(clf) + str(j) + ".png")
- fig.savefig(filename)
- j += step
- plt.close(fig)
-
-
-def distribution_data():
- """
- Plot distribution of the supervised data, i.e. the appraised iris to show how targets are distributed.
+def add_assessment_to_file(code_iris, values):
"""
- if os.path.exists(os.path.join(FOLDER_DATASETS, "data_density.csv")):
- data = pd.read_csv(os.path.join(FOLDER_DATASETS, "data_density.csv"))
- else:
- return
-
- for env in data.loc[:, ENVIRONMENT_VARIABLES].columns:
- value_env = data.loc[:, env].unique()
- distribution = [len(data[data[env] == value]) for value in value_env]
-
- if env == "geo":
- fig, ax = plt.subplots(figsize=(50, 5))
- else:
- fig, ax = plt.subplots(figsize=(10, 5))
- plt.bar(value_env, distribution)
- plt.xticks(rotation='vertical')
- plt.show()
- fig.savefig(os.path.join(FOLDER_DISTRIBUTION, "distribution-" + str(env) + ".png"))
- log.debug("Plot generated for %s", env)
+ Add an assessed IRIS to the CSV file.
+ Args:
+ code_iris: a string corresponding to the code of the IRIS (9 digits)
+ values: the values of the 6 EV that represent the environment of the assessed IRIS
-def add_assessment_to_file(code_iris, values):
+ Returns:
+ the string "okay" if the assessed IRIS has been added to the CSV file
+ """
df = pd.read_csv(FILE_MANUAL_ASSESSMENT)
- codes = df['CODE'].tolist()
+ codes = df["CODE"].tolist()
codes_lst = [str(elem) for elem in codes]
if code_iris in codes_lst:
return "iris already assessed"
@@ -418,8 +419,7 @@ def add_assessment_to_file(code_iris, values):
# adding insee indicators
indicators = model.get_indicators_list()
- indicators_to_remove = ['IRIS', 'REG', 'DEP', 'UU2010', 'COM', 'LIBCOM', 'TRIRIS', 'GRD_QUART', 'LIBIRIS',
- 'TYP_IRIS', 'MODIF_IRIS', 'LAB_IRIS', 'LIB_IRIS', 'LIB_COM', 'CODGEO', 'LIBGEO']
+ indicators_to_remove = ["IRIS", "REG", "DEP", "UU2010", "COM", "LIBCOM", "TRIRIS", "GRD_QUART", "LIBIRIS", "TYP_IRIS", "MODIF_IRIS", "LAB_IRIS", "LIB_IRIS", "LIB_COM", "CODGEO", "LIBGEO"]
for indicator in indicators_to_remove:
if indicator in indicators:
indicators.remove(indicator)
@@ -429,17 +429,6 @@ def add_assessment_to_file(code_iris, values):
iris.extend(values) # adding assessed values
cols.extend(ENVIRONMENT_VARIABLES)
- df = pd.DataFrame([iris]) #, columns=cols)
- df.to_csv(FILE_MANUAL_ASSESSMENT, mode='a', index=False, header=False) # WARNING: don't erase header in the csv file
+ df = pd.DataFrame([iris])
+ df.to_csv(FILE_MANUAL_ASSESSMENT, mode="a", index=False, header=False) # DO NOT ERASE HEADER IN THE CSV FILE
return "okay"
-
-
-if __name__ == '__main__':
- # log.info("%s", indicator_full_to_short_label("Pop 0-2 ans en 2014 (princ)"))
- # log.info("%s", indicator_short_to_full_label("P14_POP0002"))
-
- log.info(signature("RandomForestClassifier"))
-
- # distribution_data() # works only if dataset have been generated
-
- # pass
--
GitLab