In [1]:

    

import numpy as np
import networkx as nx
import gensim
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.metrics import silhouette_score
from scipy.spatial.distance import cdist, pdist
import pickle
%matplotlib inline


    

    




C:\Users\Sudarshan\Anaconda3\envs\py2\lib\site-packages\gensim\utils.py:855: UserWarning: detected Windows; aliasing chunkize to chunkize_serial
  warnings.warn("detected Windows; aliasing chunkize to chunkize_serial")

In [53]:

    

communities = []
for line in open('Fortunato_Synthetic/1000_nodes_gamma_2_beta_1_community.dat'):
    cols = line.split()
    communities.append((cols[0],cols[1]))


    

In [5]:

    

pos = nx.spring_layout(G_Fortunato)


    

In [6]:

    

nx.draw_networkx_labels(G_Fortunato,pos)
nx.draw_networkx_edges(G_Fortunato,pos,width=1.0,alpha=0.5)


    

    
Out[6]:





<matplotlib.collections.LineCollection at 0xb763780>






    

In [54]:

    

community_truth_values = [x[1] for x in communities]


    

In [55]:

    

W = pickle.load(open("emb/Fortunato_glove_1.emb", "rb" ))
G = nx.read_edgelist('Fortunato_Synthetic/1000_nodes_gamma_2_beta_1_network.dat')


    

In [56]:

    

len(W)


    

    
Out[56]:





2000

In [57]:

    

x = len(W)/2
#emb = (W[0:x,:] + W[x:,:])/2.0
emb = W[0:x,:]
emb.shape


    

    
Out[57]:





(1000L, 100L)

In [58]:

    

num_clusters = len(set(community_truth_values))
num_clusters


    

    
Out[58]:





29

In [ ]:

    




    

In [60]:

    

from collections import defaultdict
d = defaultdict(list)
kmeans = KMeans(n_clusters=num_clusters).fit(emb)
for i,j in enumerate(kmeans.labels_):
    d[j].append(str(i+1))


    

In [61]:

    

from sklearn.metrics import normalized_mutual_info_score as NMI


    

In [62]:

    

score = NMI(community_truth_values,kmeans.labels_)


    

In [63]:

    

score


    

    
Out[63]:





0.99096544345169513

In [64]:

    

def fn_cluster(cluster_list):
    cluster_map = defaultdict(int)
    i = 1
    new_labels = []
    for label in cluster_list:
        if(label not in cluster_map):
            cluster_map[label] = i
            i += 1
        new_labels.append(cluster_map[label])
    return new_labels


    

In [65]:

    

c_truth = np.asarray(fn_cluster(community_truth_values))


    

In [66]:

    

c_detect = np.asarray(fn_cluster(kmeans.labels_))


    

In [67]:

    

diff = np.abs(c_truth-c_detect)
print np.sum(diff)*1.0/len(diff)


    

    




0.367

In [70]:

    

np.count_nonzero(diff)


    

    
Out[70]:





197

In [ ]:

    




    

In [ ]: