/
enron.py
executable file
·358 lines (334 loc) · 13.4 KB
/
enron.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
#!/usr/bin/env python
import sys, os
import argparse
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import mcmc_sampler_sbm
from estimate_cluster import estimate_clustering
from sklearn.cluster import KMeans
from scipy.stats import mode
from collections import Counter
from matplotlib2tikz import save as tikz_save
## Boolean type for parser
def str2bool(v):
if v.lower() in ('yes', 'true', 't', 'y', '1'):
return True
elif v.lower() in ('no', 'false', 'f', 'n', '0'):
return False
else:
raise argparse.ArgumentTypeError('Boolean value expected.')
#########################################
## Analysis of the Enron Email Network ##
#########################################
## Import the dataset
enron = np.loadtxt('Datasets/enron_edges.txt',dtype=int,delimiter='\t')-1
###### MODEL ARGUMENTS
## PARSER to give parameter values
parser = argparse.ArgumentParser()
# Boolean variable for coclustering
parser.add_argument("-c","--coclust", type=str2bool, dest="coclust", default=False, const=False, nargs="?",\
help="Boolean variable for coclustering, default FALSE")
# Boolean variable to use second level clustering (default True)
parser.add_argument("-s","--sord", type=str2bool, dest="second_order_clustering", default=True, const=False, nargs="?",\
help="Boolean variable for second level clustering, default TRUE")
# Add options for .tex figures
parser.add_argument("-t","--tex", type=str2bool, dest="tex_figures", default=False, const=False, nargs="?",\
help="Boolean variable for .tex figures, default FALSE")
# Burnin
parser.add_argument("-B","--nburn", type=int, dest="nburn", default=25000, const=True, nargs="?",\
help="Integer: length of burnin, default 25000")
# Number of samples
parser.add_argument("-M","--nsamp", type=int, dest="nsamp", default=500000, const=True, nargs="?",\
help="Integer: length of MCMC chain after burnin, default 500000")
## Set destination folder for output
parser.add_argument("-f","--folder", type=str, dest="dest_folder", default="Results", const=True, nargs="?",\
help="String: name of the destination folder for the output files (*** the folder must exist ***)")
## Parse arguments
args = parser.parse_args()
coclust = args.coclust
second_order_clustering = args.second_order_clustering
tex_figures = args.tex_figures
nburn = args.nburn
nsamp = args.nsamp
dest_folder = args.dest_folder
# Create output directory if doesn't exist
if dest_folder != '' and not os.path.exists(dest_folder):
os.mkdir(dest_folder)
## Create the adjacency matrix
n = np.max(enron)+1
A = np.zeros((n,n))
for link in enron:
A[link[0],link[1]] = 1.0
## Construct the Gibbs sampling object
g = mcmc_sampler_sbm.mcmc_sbm(A,m=25)
## Initialise the clusters using k-means
if not coclust:
g.init_cluster(z=KMeans(n_clusters=5).fit(g.X['s'][:,:5]).labels_ if g.directed else KMeans(n_clusters=5).fit(g.X[:,:5]).labels_)
else:
g.init_cocluster(zs=KMeans(n_clusters=5).fit(g.X['s'][:,:5]).labels_,zr=KMeans(n_clusters=5).fit(g.X['r'][:,:5]).labels_)
## Average within-cluster variance
v = {}
v['s'] = np.zeros(g.m)
v['r'] = np.zeros(g.m)
for key in ['s','r']:
for k in range(g.K[key] if g.coclust else g.K):
v[key] += np.var(g.X[key][(g.z[key] if g.coclust else g.z) == k],axis=0) / (g.K[key] if g.coclust else g.K)
## Initialise d
g.init_dim(d=g.K[key] if g.coclust else g.K,delta=0.1,d_constrained=False)
## Initialise the parameters of the Gaussian distribution
g.prior_gauss_left_directed(mean0s=np.zeros(g.m),mean0r=np.zeros(g.m),Delta0s=np.diag(v['s']),Delta0r=np.diag(v['r']))
g.prior_gauss_right_directed(sigma0s=np.var(g.X['s'],axis=0),sigma0r=np.var(g.X['r'],axis=0))
## Initialise the second level clustering
if second_order_clustering:
if coclust:
g.init_group_variance_coclust(vs=range(g.K['s']),vr=range(g.K['r']))
else:
g.init_group_variance_clust(v=range(g.K),beta=1.0)
## MCMC sampler
d = []
if g.coclust:
K = {}
Ko = {}
if second_order_clustering:
H = {}
Ho = {}
for key in ['s','r']:
K[key] = []
Ko[key] = []
if second_order_clustering:
H[key] = []
Ho[key] = []
else:
K = []
Ko = []
if second_order_clustering:
H = []
Ho = []
## Posterior similarity matrix
if g.coclust:
psm = {}
for key in ['s','r']:
psm[key] = np.zeros((n,n))
else:
psm = np.zeros((n,n))
## Sampler
for s in range(nburn+nsamp):
## Print status of MCMC
if s < nburn:
sys.stdout.write("\r+++ Burnin +++ %d / %d " % (s+1,nburn))
sys.stdout.flush()
elif s == nburn:
sys.stdout.write("\n")
elif s < nburn + nsamp - 1:
sys.stdout.write("\r+++ Sweeps +++ %d / %d " % (s+1-nburn,nsamp))
sys.stdout.flush()
else:
sys.stdout.write("\r+++ Sweeps +++ %d / %d\n " % (s+1-nburn,nsamp))
## Choice of the move
if second_order_clustering:
## - with second order clustering
move = np.random.choice(['gibbs_comm','split_merge','change_dim','prop_empty','gibbs_comm_so','split_merge_so','prop_empty_so'])
else:
## - without second order clustering
move = np.random.choice(['gibbs_comm','split_merge','change_dim','prop_empty'])
if move == 'gibbs_comm':
g.gibbs_communities()
elif move == 'split_merge':
g.split_merge()
elif move == 'change_dim':
g.dimension_change(prop_step=3)
elif move == 'gibbs_comm_so':
g.gibbs_second_order()
elif move == 'split_merge_so':
g.split_merge_second_order()
elif move == 'prop_empty_so':
g.propose_empty_second_order()
else:
g.propose_empty()
## Update the parameters
if s >= nburn:
d += [g.d]
if g.coclust:
for key in ['s','r']:
K[key] += [g.K[key]]
Ko[key] += [g.Ko[key]]
if second_order_clustering:
H[key] += [g.H[key]]
Ho[key] += [np.sum(g.vk[key] > 0)]
psm[key] += np.equal.outer(g.z[key],g.z[key])
else:
K += [g.K]
Ko += [g.Ko]
if second_order_clustering:
H += [g.H]
Ho += [np.sum(g.vk > 0)]
psm += np.equal.outer(g.z,g.z)
## Convert to arrays
d = np.array(d)
if coclust:
for key in ['s','r']:
K[key] = np.array(K[key])
Ko[key] = np.array(Ko[key])
if second_order_clustering:
H[key] = np.array(H[key])
Ho[key] = np.array(Ho[key])
else:
K = np.array(K)
Ko = np.array(Ko)
if second_order_clustering:
H = np.array(H)
Ho = np.array(Ho)
## Save files
if dest_folder == '':
np.savetxt('d.txt',d,fmt='%d')
if coclust:
for key in ['s','r']:
np.savetxt('K_'+key+'.txt',K[key],fmt='%d')
np.savetxt('Ko_'+key+'.txt',Ko[key],fmt='%d')
if second_order_clustering:
np.savetxt('H_'+key+'.txt',H[key],fmt='%d')
np.savetxt('Ho_'+key+'.txt',Ho[key],fmt='%d')
np.savetxt('psm_'+key+'.txt',psm[key]/float(np.max(psm[key])),fmt='%f')
else:
np.savetxt('K.txt',K,fmt='%d')
np.savetxt('Ko.txt',Ko,fmt='%d')
if second_order_clustering:
np.savetxt('H.txt',H,fmt='%d')
np.savetxt('Ho.txt',Ho,fmt='%d')
np.savetxt('psm.txt',psm/float(np.max(psm)),fmt='%f')
else:
np.savetxt(dest_folder+'/d.txt',d,fmt='%d')
if coclust:
for key in ['s','r']:
np.savetxt(dest_folder+'/K_'+key+'.txt',K[key],fmt='%d')
np.savetxt(dest_folder+'/Ko_'+key+'.txt',Ko[key],fmt='%d')
if second_order_clustering:
np.savetxt(dest_folder+'/H_'+key+'.txt',H[key],fmt='%d')
np.savetxt(dest_folder+'/Ho_'+key+'.txt',Ho[key],fmt='%d')
np.savetxt(dest_folder+'/psm_'+key+'.txt',psm[key]/float(np.max(psm[key])),fmt='%f')
else:
np.savetxt(dest_folder+'/d.txt',d,fmt='%d')
np.savetxt(dest_folder+'/K.txt',K,fmt='%d')
np.savetxt(dest_folder+'/Ko.txt',Ko,fmt='%d')
if second_order_clustering:
np.savetxt(dest_folder+'/H.txt',H,fmt='%d')
np.savetxt(dest_folder+'/Ho.txt',Ho,fmt='%d')
np.savetxt(dest_folder+'/psm.txt',psm/float(np.max(psm)),fmt='%f')
##### Plots #####
## Scree plot
U,S,V = np.linalg.svd(A)
plt.figure()
plt.plot(np.arange(len(S))+1,S,c='black')
plt.plot(np.arange(len(S))+1,S,'.',markersize=.3,c='black')
plt.plot(mode(d)[0][0]+1,S[mode(d)[0][0]],"o",c='red')
if dest_folder == '':
if not tex_figures:
plt.savefig('scree_plot.pdf')
else:
tikz_save('scree_plot.tex')
else:
if not tex_figures:
plt.savefig(dest_folder+'/scree_plot.pdf')
else:
tikz_save(dest_folder+'/scree_plot.tex')
## Posterior barplot (unrestricted)
if coclust:
for key in ['s','r']:
plt.figure()
fig, ax = plt.subplots()
ax.bar(np.array(Counter(Ko[key]).keys())-.35,Counter(Ko[key]).values(),width=0.35,color='black',align='edge',alpha=.8,label='$K_\\varnothing$')
if second_order_clustering:
ax.bar(np.array(Counter(Ho[key]).keys()),Counter(Ho[key]).values(),width=0.35,color='gray',align='edge',alpha=.8,label='$H_\\varnothing$')
leg = ax.legend()
ax.axvline(x=mode(d)[0][0],linestyle='--',c='red')
if dest_folder == '':
if not tex_figures:
plt.savefig('posterior_barplot_unrestricted_'+key+'.pdf')
else:
tikz_save('posterior_barplot_unrestricted_'+key+'.tex')
else:
if not tex_figures:
plt.savefig(dest_folder+'/posterior_barplot_unrestricted_'+key+'.pdf')
else:
tikz_save(dest_folder+'/posterior_barplot_unrestricted_'+key+'.tex')
else:
plt.figure()
fig, ax = plt.subplots()
ax.bar(np.array(Counter(Ko).keys())-.35,Counter(Ko).values(),width=0.35,color='black',align='edge',alpha=.8,label='$K_\\varnothing$')
if second_order_clustering:
ax.bar(np.array(Counter(Ho).keys()),Counter(Ho).values(),width=0.35,color='gray',align='edge',alpha=.8,label='$H_\\varnothing$')
leg = ax.legend()
ax.axvline(x=mode(d)[0][0],linestyle='--',c='red')
if dest_folder == '':
if not tex_figures:
plt.savefig('posterior_barplot_unrestricted.pdf')
else:
tikz_save('posterior_barplot_unrestricted.tex')
else:
if not tex_figures:
plt.savefig(dest_folder+'/posterior_barplot_unrestricted.pdf')
else:
tikz_save(dest_folder+'/posterior_barplot_unrestricted.tex')
## Posterior barplot (restricted)
if coclust:
for key in ['s','r']:
plt.figure()
fig, ax = plt.subplots()
ax.bar(np.array(Counter((Ko[key])[Ko[key] >= d]).keys())-.35,Counter((Ko[key])[Ko[key] >= d]).values(),width=0.35,
color='black',align='edge',alpha=.8,label='$K_\\varnothing$')
if second_order_clustering:
ax.bar(np.array(Counter((Ho[key])[Ko[key] >= d]).keys()),Counter((Ho[key])[Ko[key] >= d]).values(),width=0.35,
color='gray',align='edge',alpha=.8,label='$H_\\varnothing$')
leg = ax.legend()
ax.axvline(x=mode(d)[0][0],linestyle='--',c='red')
if dest_folder == '':
if not tex_figures:
plt.savefig('posterior_barplot_restricted_'+key+'.pdf')
else:
tikz_save('posterior_barplot_restricted_'+key+'.tex')
else:
if not tex_figures:
plt.savefig(dest_folder+'/posterior_barplot_restricted_'+key+'.pdf')
else:
tikz_save(dest_folder+'/posterior_barplot_restricted_'+key+'.tex')
else:
plt.figure()
fig, ax = plt.subplots()
ax.bar(np.array(Counter(Ko[Ko >= d]).keys())-.35,Counter(Ko[Ko >= d]).values(),width=0.35,color='black',align='edge',alpha=.8,label='$K_\\varnothing$')
if second_order_clustering:
ax.bar(np.array(Counter(Ho[Ko >= d]).keys()),Counter(Ho[Ko >= d]).values(),width=0.35,color='gray',align='edge',alpha=.8,label='$H_\\varnothing$')
leg = ax.legend()
ax.axvline(x=mode(d)[0][0],linestyle='--',c='red')
if dest_folder == '':
if not tex_figures:
plt.savefig('posterior_barplot_restricted.pdf')
else:
tikz_save('posterior_barplot_restricted.tex')
else:
if not tex_figures:
plt.savefig(dest_folder+'/posterior_barplot_restricted.pdf')
else:
tikz_save(dest_folder+'/posterior_barplot_restricted.tex')
## MAP for clustering
if coclust:
cc_pear = {}
cc_map = {}
for key in ['s','r']:
cc_pear[key] = estimate_clustering(psm[key])
cc_map[key] = estimate_clustering(psm[key],k=mode(Ko[key])[0][0])
if dest_folder == '':
np.savetxt('pear_clusters_'+key+'.txt',cc_pear[key],fmt='%d')
np.savetxt('map_clusters_'+key+'.txt',cc_map[key],fmt='%d')
else:
np.savetxt(dest_folder+'/pear_clusters_'+key+'.txt',cc_pear[key],fmt='%d')
np.savetxt(dest_folder+'/map_clusters_'+key+'.txt',cc_map[key],fmt='%d')
else:
cc_pear = estimate_clustering(psm)
cc_map = estimate_clustering(psm,k=mode(Ko)[0][0])
if dest_folder == '':
np.savetxt('pear_clusters.txt',cc_pear,fmt='%d')
np.savetxt('map_clusters.txt',cc_map,fmt='%d')
else:
np.savetxt(dest_folder+'/pear_clusters.txt',cc_pear,fmt='%d')
np.savetxt(dest_folder+'/map_clusters.txt',cc_map,fmt='%d')