LP

embeds.py

+import numpy as np
+import networkx as nx
+import os
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import recall_score
+from sklearn.metrics import precision_score
+from sklearn.metrics import accuracy_score
+from sklearn.manifold import TSNE 
+
+from sklearn.linear_model import LogisticRegression
+
+from build_graph import node_labels
+from CNI import CNI
+
+labels_dict = node_labels()
+labels = list(labels_dict.values())
+_ , CNI_log = CNI()
+
+raw_list = [] 
+
+
+#Add CNI to original embeddings
+def embs_with_CNI():
+    
+    file = open('filename.emb', 'r')      #Enter emb file 
+    f = file.readlines()
+    del f[0]    #Removes first row in learned embeddings
+    for emb in f:
+        temp = emb.split()      #temp is a list of each row in the embedding
+        temp = [float(x) for x in temp]
+        temp[0] = int(temp[0])  #Convert the nodes from float to integer
+        get_cni = CNI_log[int(temp[0])]
+        temp.append(str(get_cni))
+        raw_list.append(temp) 
+    embeddings = np.array([i for i in raw_list]) 
+    np.savetxt('filename.txt', embeddings, fmt='%s')       #Output file / filepath
+
+
+#Unpickle .pkl embedding files from LINE and add CNI
+def unpickle():
+    raw_l = []
+    H  = nx.read_gpickle('filename.pkl')      #Enter pkl file 
+    for k, v in H.items():     
+        temp = list(v)
+        get_cni = CNI_log[int(k)]
+        temp.append(get_cni)
+        raw_l.append([int(k), *temp])
+    embeddings = np.array([i for i in raw_l]) 
+    np.savetxt('filename.txt', embeddings, fmt='%s')       #Output file / filepath
+
+#unpickle()
+
+
+def read_EmbsCNI(file):
+    for path in file:
+        extension = os.path.splitext(path)[-1].lower()
+        filebody = os.path.splitext(path)[0].upper()
+        if extension == '.emb':
+            name = "GRAPH WITH INHERENT LABELS // " + filebody
+            np_emb = np.loadtxt(open(file[-1]), skiprows = 1)
+            nodes = [int(x[:1]) for x in np_emb]
+            features = [labels_dict[n] for n in nodes] 
+            nodes_emb = [x[1:] for x in np_emb]     
+            return (nodes, features, nodes_emb, name)
+            #model = TSNE(n_components=1)
+            #two_dim = model.fit_transform(nodes_emb)
+            #return (features, two_dim)
+        elif extension == '.txt':
+            name = "GRAPH WITH CNI ENCODING // "  + filebody
+            np_embCNI = np.loadtxt(open(file[-1]))
+            np.random.shuffle(np_embCNI)
+            nodes = [int(x[:1]) for x in np_embCNI] 
+            enc_feature = [labels_dict[n] for n in nodes] 
+            nodes_emb = [x[1:] for x in np_embCNI]
+            return (nodes, enc_feature, nodes_emb, name)
+        elif extension == '.pkl':
+            name = "GRAPH WITH INHERENT LABELS // " + filebody
+            H  = nx.read_gpickle(file[-1])
+            nodes = [int(k) for k, _ in H.items()] 
+            features = [labels_dict[n] for n in nodes] 
+            nodes_emb = [v for _, v in H.items()] 
+            return (nodes, features, nodes_emb, name)
+        else:
+            print("file does not exist")
+
+
+def LogReg(features, node_emb, name):
+    
+    xtrain, xtest, ytrain, ytest = train_test_split(node_emb, features, 
+                                                    test_size = 0.3, random_state = 25)
+    model = LogisticRegression(class_weight='balanced')
+    model.fit(xtrain, ytrain)
+    LogisticRegression(C=1.0, class_weight='balanced', dual=False,
+              fit_intercept=True, intercept_scaling=1, max_iter=1500,
+              multi_class='ovr', n_jobs=1, penalty='l2', random_state=None,
+              solver='liblinear', tol=0.0001, verbose=0, warm_start=False)
+     
+    predicted = model.predict(xtest)
+     
+    accu = accuracy_score(ytest, predicted)
+    prs = precision_score(ytest, predicted, average='weighted')
+    rcs = recall_score(ytest, predicted, average='weighted')
+    print('>>>', name, 'LogReg', end='\n')
+    print('Link Prediction Accuracy Score: ', accu, end='\n') 
+    print('Precision Score: ', prs, end='\n')
+    print('Recall Score: ', rcs)
+    print()
+
+_, features, node_emb, name = read_EmbsCNI(['filename.emb'])      #Read emb or txt file
+#LogReg(features, node_emb, name)
+
+