In [26]:

    

import os
import networkx as nx
import pickle
from sklearn.metrics import normalized_mutual_info_score as NMI
from sklearn.cluster import SpectralClustering
import numpy as np


    

In [3]:

    

path = "../../Initial_Test/Generated Graphs/"
directories = os.listdir(path)
for x in directories :
    exe = "mkdir "+x
    os.system(exe)


    

In [53]:

    

for x in directories :
    a = os.listdir(path+x)
    print x
    for m in a :
        if "network" in m:
            communities = {}
            readfile = os.path.abspath(path+x+"\\"+m)
            writefile = os.path.abspath(x+"\\"+m)
            writefile = writefile.replace("network","community")
            G = nx.read_edgelist(readfile)
            adj = nx.adjacency_matrix(G)
            commfile = m.replace("network","community")
            rf2 = os.path.abspath(path+x+"\\"+commfile)
            with open(rf2) as handle :
                temp = handle.readlines()
            number_comm = len(set([a.split()[1] for a in temp])) #Finding the true number of communities
            clus = SpectralClustering(n_clusters=number_comm)
            out = clus.fit_predict(adj)
            graph_nodes = G.nodes()
            for d in range(len(graph_nodes)) :
                communities[int(graph_nodes[d])] = out[d]
            with open(writefile,'wb') as handle :
                for nodes in range(1,1001) :
                    temp = str(nodes) + "\t" + str(communities[nodes])
                    handle.write(temp+"\n")


    

    




mu_0.05_N_1000

In [19]:

    

urange = [0.05*i for i in range(20)]


    

In [54]:

    

ugraph = []
for x in urange:
    values = []
    truth_comm = "../../Initial_Test/Generated Graphs/"
    directory = "mu_"+str(x)+"_N_1000/"
    for y in range(1,51):
        true_path = truth_comm + directory + "community_run_"+str(y)+".dat"
        with open(true_path) as handle :
            temp = handle.readlines()
        tc = [a.split()[1] for a in temp]
        spectral_path = directory + "community_run_"+str(y)+".dat"
        with open(spectral_path) as handle :
            temp = handle.readlines()
        lc = [a.split()[1] for a in temp]
        val = NMI(tc,lc)
        values.append(val)
    ugraph.append(sum(values)/50.0)


    

In [55]:

    

with open("spectral_scores.pickle",'wb') as handle:
    pickle.dump(ugraph,handle)


    

In [56]:

    

print ugraph


    

    




[0.20202325739106627]

In [ ]: