Made a nice script for baseline model training

train_baseline.py

+# coding = utf-8
+# -*- coding: utf-8 -*-
+
+import os
+import argparse
+
+# BASIC
+import pandas as pd
+import numpy as np
+
+# ML
+# MACHINE LEARNING
+from sklearn.feature_extraction.text import CountVectorizer
+from sklearn.feature_extraction.text import TfidfTransformer
+from sklearn.pipeline import Pipeline
+from sklearn.naive_bayes import MultinomialNB
+from sklearn.svm import SVC
+from sklearn.linear_model import SGDClassifier
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.neighbors import KNeighborsClassifier
+from sklearn.ensemble import RandomForestClassifier
+
+from sklearn.model_selection import GridSearchCV
+
+# ML HELPERS
+from sklearn import preprocessing
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import classification_report
+
+from glob import glob
+from joblib import dump,load
+
+#PROGRESS BAR
+from tqdm import tqdm
+
+from ngram import NGram
+
+parser = argparse.ArgumentParser()
+parser.add_argument("dataset_name")
+parser.add_argument("inclusion_fn")
+parser.add_argument("adjacency_fn")
+parser.add_argument("cooc_fn")
+
+args = parser.parse_args()
+
+DATASET_NAME = args.dataset_name
+I_FN = args.inclusion_fn
+A_FN = args.adjacency_fn
+C_FN = args.cooc_fn
+
+OUTPUT_DIR = "outputs"
+
+verbose = False
+
+for fn in [I_FN,A_FN,C_FN]:
+    if not os.path.exists(fn):
+        raise FileNotFoundError("{0} does not exists !".format(fn))
+
+
+
+classifier_dict = {
+    "naive-bayes":MultinomialNB(),
+    "svm":SVC(),
+    "sgd":SGDClassifier(),
+    "knn":KNeighborsClassifier(),
+    "decision-tree": DecisionTreeClassifier(),
+    "random-forest":RandomForestClassifier()
+}
+
+parameters = {
+    "naive-bayes":[{"alpha":[0,1]}],
+    "svm":[{"kernel":["rbf","poly"], 'gamma': [1e-1,1e-2,1e-3, 1,10,100]}],
+    "sgd":[{"penalty":["l1","l2"],"loss":["hinge","modified_huber","log"]}],
+    "knn":[{"n_neighbors":list(range(4,8)),"p":[1,2]}],
+    "decision-tree": [{"criterion":["gini","entropy"]}],
+    "random-forest":[{"criterion":["gini","entropy"],"n_estimators":[10,50,100]}]
+}
+
+combinaison = [
+    [I_FN,C_FN,A_FN],
+    [I_FN,C_FN],
+    [C_FN],
+    [C_FN,A_FN]
+]
+combinaison_label = [
+    "PIC",
+    "IC",
+    "C",
+    "PC"
+]
+
+for ix, comb in enumerate(combinaison):
+    df = pd.concat([pd.read_csv(fn,sep="\t").head(500) for fn in comb])
+
+    index = NGram(n=4)
+    data_vectorizer = Pipeline([
+        ('vect', CountVectorizer(tokenizer=index.split)),
+        ('tfidf', TfidfTransformer()),
+    ])
+
+    X_train,y_train = (df[df.split == "train"].toponym + " " + df[df.split == "train"].toponym_context).values, df[df.split == "train"].hp_split
+    X_test,y_test = (df[df.split == "test"].toponym + " " + df[df.split == "test"].toponym_context).values, df[df.split == "test"].hp_split
+    data_vectorizer.fit((df.toponym + " " + df.toponym_context).values)
+
+    dump(data_vectorizer,"{2}/{0}_{1}_vectorizer.pkl".format(DATASET_NAME,combinaison_label[ix],OUTPUT_DIR))
+
+    X_train = data_vectorizer.transform(X_train)
+    X_test = data_vectorizer.transform(X_test)
+
+    for CLASSIFIER in tqdm(classifier_dict):
+        if verbose : print("TRAIN AND EVAL {0}".format(CLASSIFIER))
+        clf = GridSearchCV(
+                classifier_dict[CLASSIFIER], parameters[CLASSIFIER], scoring='f1_weighted',n_jobs=-1
+            )
+        clf.fit(X_train, y_train)
+        if verbose : print("Best Parameters : ",clf.best_params_)
+        y_pred = clf.best_estimator_.predict(X_test)
+        if verbose : print(classification_report(y_test,y_pred))
+        dump(clf.best_estimator_,"{0}/{1}_{2}_{3}.pkl".format(OUTPUT_DIR,DATASET_NAME,combinaison_label[ix],CLASSIFIER))
\ No newline at end of file