/
peoplefinder.py
443 lines (392 loc) · 18.9 KB
/
peoplefinder.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
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
# -*- coding: utf-8 -*-
import os
import time
import pp # parallel python
import jieba
import pickle
import selenium # parse dynamic web pages
import numpy
import scipy
import munkres
import sys
sys.path.extend(['mail', 'social/linkedin', 'social/renren'])
from mailcleaner import MailCleaner
from visualgraph import VisualGraph
from renren import RenRen
PROJECT_ROOT = os.path.split(os.path.realpath(__file__))[0]
class PeopleFinder(object):
def __init__(self, mail_handler, social_handler):
self.mail_handler = mail_handler
self.social_handler = social_handler
self.profile_results = {}
def create_email_network(self, local=False, show_and_save=False):
if not local:
self.email_mapping_table, self.email_contact_table = self.mail_handler.clean_emailaddrs(\
mc.DEFAULT_STOP_PATTERN)
# save data
self.save_data([self.email_mapping_table, self.email_contact_table], 'email')
else:
self.email_mapping_table, self.email_contact_table = self.load_data('email')
if show_and_save:
vg = VisualGraph()
vg.import_data(self.email_contact_table)
vg.save_graph('tmp_mail.png')
return self.email_mapping_table, self.email_contact_table
def create_social_network(self, mapping_table):
top_num = 10
first_circle_candidates = {}
social_friend_table = {} # friend list
driver = selenium.webdriver.Firefox() # assume you have firefox on your local computer
# login first
driver.get('http://renren.com')
username = driver.find_element_by_id("email")
password = driver.find_element_by_id("password")
username.send_keys(self.social_handler.email)
password.send_keys(self.social_handler.password)
driver.find_element_by_id("login").click()
time.sleep(3) # import time delay
for emailaddr, username in mapping_table:
candidates_table = self.social_handler.search_profiles(emailaddr, top_num, driver)
# candidates_table = {}
if not candidates_table:
search_name = username and username or emailaddr.split('@')[0]
candidates_table = self.social_handler.search_profiles(search_name, top_num, driver)
first_circle_candidates.update(candidates_table)
driver.close()
social_mapping_table = first_circle_candidates.copy()
for each_uid in first_circle_candidates.keys():
all_friends = self.social_handler.get_friends(each_uid)
social_mapping_table.update(all_friends)
social_friend_table.update({each_uid: set(all_friends.keys())})
for each_friend in all_friends.keys():
try:
social_friend_table[each_friend].update([each_uid])
except:
social_friend_table.update({each_friend: set([each_uid])})
return social_mapping_table, social_friend_table
def create_social_network_pp(self, mapping_table, local=False, show_and_save=False):
if not local:
self.social_mapping_table = {}
self.social_friend_table = {}
mapping_table = mapping_table.items()
# using parallel computing
batch_num = 8
task_num = len(mapping_table)
batch_size = task_num/batch_num
job_server = pp.Server()# require parallel python
for index in range(0, batch_num):
job = job_server.submit(func=self.create_social_network, \
args=(mapping_table[index*batch_size:(index+1)*batch_size],), \
depfuncs=(), modules=('selenium', 'time'), callback=self.merge_social_network)
job_server.wait()
print "%s tasks done !"%(batch_num*batch_size)
if task_num - batch_num*batch_size != 0:
job = job_server.submit(func=self.create_social_network, \
args=(mapping_table[batch_num*batch_size:task_num],), \
depfuncs=(), modules=('selenium', 'time'), callback=self.merge_social_network)
job_server.wait()
print "%s tasks done !"%task_num
# save data
self.save_data([self.social_mapping_table, self.social_friend_table], 'social')
else:
self.social_mapping_table, self.social_friend_table = self.load_data('social')
if show_and_save:
pass
return self.social_mapping_table, self.social_friend_table
def merge_social_network(self, result):
self.social_mapping_table.update(result[0])
for k, v in result[1].iteritems():
try:
self.social_friend_table[k].update(v)
except:
self.social_friend_table.update({k:v})
def save_data(self, data, data_type):
"""
Save your data in local environment
"""
path_root = os.path.join(PROJECT_ROOT, 'data', data_type)
if not os.path.exists(path_root):
os.makedirs(path_root)
try:
with open(os.path.join(path_root, 'mapping_table.dat'), 'w') as f1, \
open(os.path.join(path_root, 'contact_table.dat'), 'w') as f2:
mapping_table = pickle.dumps(data[0])
contact_table = pickle.dumps(data[1])
f1.write(mapping_table)
f2.write(contact_table)
except Exception, e:
print e
return False
return True
def load_data(self, data_type):
path_root = os.path.join(PROJECT_ROOT, 'data', data_type)
try:
with open(os.path.join(path_root, 'mapping_table.dat'), 'r') as f1, \
open(os.path.join(path_root, 'contact_table.dat'), 'r') as f2:
mapping_table = pickle.loads(f1.read())
contact_table = pickle.loads(f2.read())
except Exception, e:
print e
return False
return [mapping_table, contact_table]
def save_results(self, data, data_type):
"""
Save your results in local environment
"""
path_root = os.path.join(PROJECT_ROOT, 'data/results')
if not os.path.exists(path_root):
os.makedirs(path_root)
try:
with open(os.path.join(path_root, '%s.dat'%data_type), 'w') as f:
results = pickle.dumps(data)
f.write(results)
except Exception, e:
print e
return False
return True
def load_results(self, data_type):
path_root = os.path.join(PROJECT_ROOT, 'data/results')
try:
with open(os.path.join(path_root, '%s.dat'%data_type), 'r') as f:
results = pickle.loads(f.read())
except Exception, e:
print e
return False
return results
def merge_results(self, result):
self.recommend_list.update(result)
def run_pp(self, method, top_num=10):
email_mapping_table = self.email_mapping_table.items()
if method == 'graph':
self.recommend_list = self.run(method, email_mapping_table, top_num)
else:
self.recommend_list = {}
# using parallel computing
batch_num = 8
task_num = len(email_mapping_table)
batch_size = task_num/batch_num
job_server = pp.Server()# require parallel python
for index in range(0, batch_num):
job = job_server.submit(func=self.run, \
args=(method, email_mapping_table[index*batch_size:(index+1)*batch_size], top_num),\
depfuncs=(), modules=('jieba',), callback=self.merge_results)
job_server.wait()
print "%s tasks done !"%(batch_num*batch_size)
if task_num - batch_num*batch_size != 0:
job = job_server.submit(func=self.run, \
args=(method, email_mapping_table[batch_num*batch_size:task_num], top_num),\
depfuncs=(), modules=('jieba',), callback=self.merge_results)
job_server.wait()
print "%s tasks done !"%task_num
self.save_results(self.recommend_list, method)
def run(self, method, email_mapping_table, top_num=10):
if method == 'graph':
self.email_num = len(self.email_mapping_table)
self.social_num = len(self.social_mapping_table)
self.email_index2uid = self.email_mapping_table.keys()
self.social_index2uid = self.social_mapping_table.keys()
self.email_uid2index = dict(zip(self.email_index2uid, range(self.email_num)))
self.social_uid2index = dict(zip(self.social_index2uid, range(self.social_num)))
recommend_list = self.do_recommend(method, top_num)
else:
recommend_list = {}
for each_email_uid, each_email_name in email_mapping_table:
candidates = self.do_recommend(method, top_num, each_email_uid)
recommend_list[each_email_uid] = candidates
print '%s done!'%each_email_uid
# self.save_results(recommend_list, method)
return recommend_list
def do_recommend(self, method, top_num, email_uid=''):
candidates = {}
if method == 'profile':
email_pf = self.email_mapping_table[email_uid] and self.email_mapping_table[email_uid] or email_uid.split('@')[0]
for each_social_uid, each_social_pf in self.social_mapping_table.iteritems():
sim = self.calc_profile_sim(email_pf, each_social_pf)
candidates.update({each_social_uid: sim})
elif method == 'graph':
self.calc_graph_sim()
recommend_list = {}
for each_email_index in range(self.email_num):
email_uid = self.email_index2uid[each_email_index]
candidates = zip(self.social_index2uid, self.graph_sim_matrix[each_email_index])
candidates = sorted(candidates, key=lambda d:d[1], reverse=True)
recommend_list[email_uid] = candidates[:top_num]
return recommend_list
elif method == 'overlap':
for each_social_uid in self.social_mapping_table.keys():
sim = self.calc_entry_sim_overlap(email_uid, each_social_uid)
candidates.update({each_social_uid: sim})
else:
pass
candidates = sorted(candidates.iteritems(), key=lambda d:d[1], reverse=True)
return candidates[:top_num]
def calc_graph_sim(self, threshold_list = [0.1, 0.01, 0.001, 0.0001, 0.00001]):
self.graph_sim_matrix = numpy.ones((self.email_num, self.social_num)) # initial state
self.profile_sim_matrix = -numpy.ones((self.email_num, self.social_num)) # initial state
changes = 1.0
for each_threshold in threshold_list:
while True:
if changes <= each_threshold:
break
changes = self.itcalc_graph_sim_matrix()
# for test threshold
try:
numpy.save('graph_sim_mat_c%s.npy'%changes, self.graph_sim_matrix)
except Exception, e:
print e
pass
print 'changes: %s costs %ss'%(changes, (time.time()-t0))
print 'threshold: %s costs %ss'%(each_threshold, (time.time()-t0))
recommend_list = {}
for each_email_index in range(self.email_num):
email_uid = self.email_index2uid[each_email_index]
candidates = zip(self.social_index2uid, self.graph_sim_matrix[each_email_index])
candidates = sorted(candidates, key=lambda d:d[1], reverse=True)
recommend_list[email_uid] = candidates[:10]
self.save_results(recommend_list, 'graph_%s'%each_threshold)
return self.graph_sim_matrix
def itcalc_graph_sim_matrix(self, threshold=0.1):
valid_count = 0
tmp_sim_matrix = scipy.sparse.lil_matrix((self.email_num, self.social_num))
for each_email_uid, each_email_uname in self.email_mapping_table.iteritems():
if not self.email_contact_table[each_email_uid]:
continue
email_index = self.email_uid2index[each_email_uid]
for each_social_uid, each_social_friends in self.social_friend_table.iteritems():
if each_social_friends:
social_index = self.social_uid2index[each_social_uid]
if self.profile_sim_matrix[email_index, social_index] >= threshold:
pass
elif self.profile_sim_matrix[email_index, social_index] == -1:
email_pf = each_email_uname and each_email_uname or each_email_uid.split('@')[0]
profile_sim = self.calc_profile_sim(email_pf, self.social_mapping_table[each_social_uid])
self.profile_sim_matrix[email_index, social_index] = profile_sim
if profile_sim < threshold:
continue
else:
continue
sim = self.fuzzy_jaccard_sim(each_email_uid, each_social_uid)
tmp_sim_matrix[email_index, social_index] = sim
valid_count += 1
print '%s done!'%each_email_uid
changes = valid_count and numpy.sum(abs(self.graph_sim_matrix - tmp_sim_matrix))/valid_count or 0.0
self.graph_sim_matrix = tmp_sim_matrix.toarray()
print 'valid_count:%s'%valid_count
return changes
def fuzzy_jaccard_sim(self, email_uid, social_uid):
email_index_list = [self.email_uid2index[each_uid] for each_uid in self.email_contact_table[email_uid] if each_uid != email_uid]
social_index_list = [self.social_uid2index[each_uid] for each_uid in self.social_friend_table[social_uid]]
neighboring_matrix = self.graph_sim_matrix[email_index_list,:][:,social_index_list]
# transpose the matrix if row number > col number
row, col = neighboring_matrix.shape
if row > col:
neighboring_matrix = neighboring_matrix.transpose()
mk = munkres.Munkres()
try:
indexes = mk.compute(-neighboring_matrix)
except Exception, e:
print e
# import pdb;pdb.set_trace()
return 0.0
fuzzy_intersection = 0.0
for row, col in indexes:
fuzzy_intersection += neighboring_matrix[row, col]
fuzzy_jaccard = fuzzy_intersection/(self.email_num+self.social_num-fuzzy_intersection)
return fuzzy_jaccard
def calc_entry_sim_overlap(self, email_uid, social_uid):
RATIO = 0.5
# neighborhood_optimal_match = {}
overlap_score = 0.0
total_score = 0.0
for each_neighbor in self.email_contact_table[email_uid]:
optimal_match = self.get_optimal_socials([each_neighbor, \
self.email_mapping_table[each_neighbor]])
# neighborhood_optimal_match[optimal_match[0]] = optimal_match[1]
total_score += optimal_match[1]
if optimal_match[0] in self.social_friend_table[social_uid]:
overlap_score += optimal_match[1]
# Method 1
sim_overlap_a = total_score and overlap_score/total_score or 0.0
# Method 2
count = len(self.social_friend_table[social_uid])
sim_overlap_b = count and overlap_score/count or 0.0
sim_overlap = sim_overlap_a*RATIO + sim_overlap_b*(1-RATIO)
return sim_overlap
def get_optimal_socials(self, email_entry, local=True):
if local:
return self.profile_results[email_entry[0]][0]
result_list = {}
email_pf = email_entry[1] and email_entry[1] or email_entry[0].split('@')[0]
for each_uid, each_uname in self.social_mapping_table.iteritems():
sim = self.calc_profile_sim(email_pf, each_uname)
result_list.update({each_uid: sim})
optimal_match = sorted(result_list.iteritems(), key=lambda d:d[1], reverse=True)[:1]
return optimal_match
def calc_profile_sim(self, email_pf, social_pf):
sim = self.calc_string_sim(email_pf, social_pf)
return sim
def calc_string_sim(self, a, b):
try:
a_seg_list = [x for x in jieba.cut(a, cut_all=False)]
b_seg_list = [x for x in jieba.cut(b, cut_all=False)]
except Exception, e:
print e
assert False
sim = self.jaccard_sim(a_seg_list, b_seg_list)
return sim
def jaccard_sim(self, a, b): # remove repeated words
a, b = set(a), set(b)
union_count = float(len(a | b))
sim = union_count and len(a&b)/union_count or 0
return sim
def calc_profile_sim_matrix(self):
real_mapping = {}
try:
with open('data/results/real_mapping.dat', 'r') as f:
for line in f:
content = line.split(',')
if len(content) == 3:
real_mapping[content[0]] = content[2].replace('\n', '')
f.close()
except Exception, e:
print e
return False
email_index2uid = self.email_mapping_table.keys()
social_index2uid = self.social_mapping_table.keys()
email_uid2index = dict(zip(email_index2uid, range(len(self.email_mapping_table))))
social_uid2index = dict(zip(social_index2uid, range(len(self.social_mapping_table))))
profile_sim_matrix = numpy.zeros((len(real_mapping), len(self.social_mapping_table)))
index = 0
for each_email_uid in real_mapping.keys():
email_pf = self.email_mapping_table[each_email_uid] and self.email_mapping_table[each_email_uid] or each_email_uid.split('@')[0]
for each_social_uid, each_social_pf in self.social_mapping_table.iteritems():
sim = self.calc_profile_sim(email_pf, each_social_pf)
if sim >= 0.1:
profile_sim_matrix[index, social_uid2index[each_social_uid]] = sim
index += 1
print '%s done!'%each_email_uid
try:
numpy.save('profile_sim_matrix_part.npy', profile_sim_matrix)
except Exception, e:
print e
t0 = 0
if __name__ == '__main__':
mbox_path = ['data/email/qq_mail.mbox']
email = "XXXXXXXXX"
passwd = 'XXXXXXXXX'
mc = MailCleaner(mbox_path)
renren = RenRen(email, passwd)
# renren = None
pf = PeopleFinder(mc, renren)
ef = pf.create_email_network(local=True)
V, E = pf.create_social_network_pp(ef[0], local=True)
# t0 = time.time()
# pf.run_pp('profile', 10)
# pf.calc_profile_sim_matrix()
# print 'costs %ss'%(time.time()-t0)
# vg = VisualGraph(name='social network')
# vg.import_data(E)
# vg.save_graph('social network')
# vg.save_histogram('social histogram')
# optimal_match = pf.get_optimal_socials(['email', 'name'])
# print optimal_match