-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathRecommenderSystem.py
379 lines (282 loc) · 14.1 KB
/
RecommenderSystem.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
# coding: utf-8
# In[136]:
from __future__ import division
import pandas as pd
import numpy as np
import os
from scipy import cluster
import pickle
# import plotly.plotly as py
# import matplotlib.pyplot as plt
# %matplotlib inline
import math
from sklearn import preprocessing
import get_metro_features as gmf
import get_city_proportion as gcp
import plot_and_split as pas
import datetime
import scipy
from scipy.sparse import csc_matrix
# In[155]:
# cur_dir = os.getcwd()
# dataset_dir = cur_dir
# ext = ''
cur_dir = '~/Desktop/dataset'
dataset_dir = cur_dir
ext = ''
CITYNAME = 'LasVegas'
skip = False
if skip:
variables = pickle.load(open('data/variables.pck', 'rb'))
cities=variables['cities']
connections=variables['connections']
user_res=variables['user_res']
major_businesses=variables['major_businesses']
CITYNAME=variables['CITYNAME']
else:
# In[156]:
users = pd.read_csv(dataset_dir + '/user' + ext + '.csv')
reviews = pd.read_csv(dataset_dir + '/review' + ext + '.csv')
businesses = pd.read_csv(dataset_dir + '/business' + ext + '.csv')
# In[157]:
major_businesses, state_num = pas.eliminate_minor_states(businesses)
# Manually define boundaries because that's the easiest way to do this
boundaries = dict(AZ=dict(bot=31.30, top=37.0, left=-115.0, right=-109),
NC=dict(bot=33.8, top=36.8, left=-84.4, right=-75.1),
PA=dict(bot=39, top=42.3, left=-80.7, right=-74.6),
NV=dict(bot=34.9, top=42.07, left=-120.1, right=-114.02))
# One odd point this doesn't clean for NV because of diagonal, but whatever, not worth a more advanced function
for state in state_num:
if state not in boundaries.keys():
# All other states are clean
continue
# Filter the one's that do need to be filtered
major_businesses = major_businesses[(major_businesses.state != state) | (major_businesses['latitude'] > boundaries[state]['bot'])]
major_businesses = major_businesses[(major_businesses.state != state) | (major_businesses['latitude'] < boundaries[state]['top'])]
major_businesses = major_businesses[(major_businesses.state != state) | (major_businesses['longitude'] > boundaries[state]['left'])]
major_businesses = major_businesses[(major_businesses.state != state) | (major_businesses['longitude'] < boundaries[state]['right'])]
# Also NC and SC should be combined, because its the same area
major_businesses.loc[major_businesses.state=='SC', 'state'] = 'NC'
# In[158]:
states = set(major_businesses['state'])
print(states)
init = np.zeros((len(states), 2))
for i, state in enumerate(states):
init_pt = major_businesses[major_businesses['state'] == state].sample(1)
init[i, 0] = init_pt['latitude']
init[i, 1] = init_pt['longitude']
# It looks like we have one NaN in lat/lng
major_businesses.dropna(subset=['latitude', 'longitude'], inplace=True)
clusters = pas.cluster_cities(major_businesses, k=11, iter=500, init=init)
# In[159]:
major_businesses = major_businesses.assign(metro_area=pd.Series(clusters[1]).values)
print('major_businesses shape: {}'.format(major_businesses.shape))
major_businesses = major_businesses[major_businesses['review_count'] > 30] #impose 30 review minimum for business inclusion
print('major_businesses shape: {}'.format(major_businesses.shape))
# In[160]:
# b = major_businesses.keys()
# nonA = [ba for ba in b if 'Attributes' not in ba and 'attributes' not in ba and 'hours' not in ba]
# major_businesses2 = major_businesses[nonA]
# # major_businesses2.groupby('metro_area')
# major_businesses2 = major_businesses2.rename(columns={'review_count':'biz_review_count'})
# In[161]:
join_date = pd.to_datetime(users['yelping_since']).dt.date
now_date = datetime.date(2017, 12, 1)
users['weeks_on_yelp'] = (now_date - join_date).dt.days / 7
print('user shape: {}'.format(users.shape))
# Eliminate users with less than 20 reviews
users = users[users.review_count >= 20]
print('user shape after cutting: {}'.format(users.shape))
# In[162]:
connections = gmf.link_for_metro(users, major_businesses, reviews)
# This gives us percentage of reviews in each metro area
perc_metro = gmf.calc_perc_metro(connections)
# print(perc_metro)
# In[163]:
num_visited = gmf.num_metros_visited(perc_metro)
weeks_metro = gmf.calc_num_weeks_metro(num_visited)
# In[164]:
# review_metro = gmf.reviews_per_week_per_metro(weeks_metro)
user_features = gmf.define_user_features(users, weeks_metro, clusters[0].shape[0])
# user_features['num_metros_visited'] = perc_metro.groupby('user_id')['num_metros_visited'].nunique()
# user_features = pd.merge(user_features, dnew, on='user_id', how='left')
user_features.fillna(0, inplace=True)
user_features = user_features.groupby('user_id').max()
user_features = user_features.reset_index()
# In[165]:
user_f = np.array(user_features[['m0_percent', # removed 'num_metros_visited', 'weeks_on_yelp',
'm0_weeks', 'm1_percent', 'm1_weeks', 'm2_percent', 'm2_weeks',
'm3_percent', 'm3_weeks', 'm4_percent', 'm4_weeks', 'm5_percent',
'm5_weeks', 'm6_percent', 'm6_weeks', 'm7_percent', 'm7_weeks',
'm8_percent', 'm8_weeks', 'm9_percent', 'm9_weeks', 'm10_percent',
'm10_weeks']])
user_f = user_f.astype(float)
initialization = np.zeros((11, user_f.shape[1]))
for i in range(11):
initialization[i, 2*i] = 1
print(np.max(initialization, axis=1))
# print(initialization)
# print(np.sum(user_f, axis=1)[0:20])
# # print(user_features[0:5])
# # user_fnorm = preprocessing.normalize(user_f, norm='l1', axis=0, copy=True, return_norm=False)
user_clustering = cluster.vq.kmeans2(user_f, initialization, iter=600, minit='matrix')
user_res = user_features.copy()
user_res['group'] = user_clustering[1]
print(np.max(user_clustering[0], axis=1))
print(user_res.keys())
# In[166]:
user_res['cluster_dist'] = np.zeros(user_res.shape[0])
cluster_keys = [k for k in user_res.keys() if k not in ['user_id', 'group', 'cluster_dist', 'num_metros_visited', 'weeks_on_yelp']]
for m in range(1, clusters[0].shape[0]):
test_group = user_res.loc[user_res.group == m, cluster_keys]
# print(test_group)
user_res.loc[user_res.group == m, 'cluster_dist'] = np.linalg.norm(user_res.loc[user_res.group == m, cluster_keys] -
user_clustering[0][m], axis=1)
# Cut out users with less than 20 reviews
# In[167]:
city_centers = dict(
Toronto=(43.66, -79.58),
LasVegas=(36.0, -115.0),
Edinburgh=(55.0, -3.19),
Pittsburgh=(40.12, -80.12),
Madison=(43.07, -89.38),
Champaign=(40.08, -88.29),
Stuttgart=(48.71, 9.23),
Phoenix=(33.39, -111.91),
Cleveland=(41.18, -81.49),
Charlotte=(35.25, -80.79),
Montreal=(45.58, -73.54)
)
def closest_value(compare, lat, lng):
mindist = 1e6
mindex = -1
for k in compare:
temp = np.linalg.norm([compare[k][0] - lat, compare[k][1] - lng])
if temp < mindist:
mindex = k
mindist = temp
# print('new mindist = {}'.format(mindist))
return mindex
cities = dict()
city_bs = dict()
for i in range(clusters[0].shape[0]):
city_locals = user_res[user_res['group'] == i]
connection_locals = connections[connections['user_id'].isin(city_locals['user_id'])]
mean_lat, mean_lng = np.mean(connection_locals['latitude']), np.mean(connection_locals['longitude'])
# print(mean_lat, mean_lng)
city = closest_value(city_centers, mean_lat, mean_lng)
cities[city] = i
print('User city {} = {}\t {:.2f}, {:.2f}'.format(city, i, mean_lat, mean_lng))
# Now do businesses
business_locals = major_businesses[major_businesses['metro_area'] == i]
mean_lat, mean_lng = np.mean(business_locals['latitude']), np.mean(business_locals['longitude'])
city = closest_value(city_centers, mean_lat, mean_lng)
city_bs[city] = i
print('Biz city {} = {}\t {:.2f}, {:.2f}'.format(city, i, mean_lat, mean_lng))
print('===================================')
# print(cities)
# In[168]:
city_centers = """
#: \tlat, lng =\tCity
0:\t43.60, 79.50=\tToronto
1:\t36.0, -115.0=\tLas Vegas
2:\t55.0, -3.19=\tEdinburgh
3:\t43.0, -79.9=\tPittsburgh
4:\t43.07, -89.38=\tMadison
5:\t40.08, -88.29=\tChampaign
6:\t48.71, 9.23=\tStuttgart
7:\t33.39, -111.91=\tPhoenix
8:\t41.18, -81.49=\tCleveland
9:\t35.25, -80.79=\tCharlotte
10:\t45.58, -73.54=\tMontreal
"""
# print(city_centers)
variables = dict(
cities=cities,
connections=connections,
user_res=user_res,
major_businesses=major_businesses,
CITYNAME=CITYNAME
)
pickle.dump(variables, open('data/variables.pck', 'wb'))
def get_city_array(city_index, connections, user_res, businesses, savename=None):
local_users = user_res[user_res['group'] == city_index]
local_bizes = businesses[businesses['metro_area'] == city_index]
local_connections = connections[connections['user_id'].isin(local_users['user_id'])]
local_connections = local_connections[local_connections['business_id'].isin(local_bizes['business_id'])]
# For tourist, get new users but same businesses
tourist_users = user_res[user_res['group'] != city_index]
tourist_bizes = businesses[businesses['metro_area'] == city_index]
tourist_connections = connections[connections['user_id'].isin(tourist_users['user_id'])]
tourist_connections = tourist_connections[tourist_connections['business_id'].isin(tourist_bizes['business_id'])]
biz_ids = pd.merge(local_connections[['business_id']], tourist_connections[['business_id']], on=['business_id'], how='inner')
local_connections = local_connections[local_connections['business_id'].isin(biz_ids['business_id'])]
tourist_connections = tourist_connections[tourist_connections['business_id'].isin(biz_ids['business_id'])]
# Now set up local array
local_users = set(local_connections.user_id)
local_businesses = set(local_connections.business_id)
local_data = local_connections['stars'].tolist()
local_col = local_connections.user_id.astype('category', categories=local_users).cat.codes
local_row = local_connections.business_id.astype('category', categories=local_businesses).cat.codes
assert(local_col[local_col<0].shape[0] == 0)
assert(local_row[local_row<0].shape[0] == 0)
local_sparse_matrix = csc_matrix((local_data, (local_row, local_col)),
shape=(len(local_businesses), len(local_users)))
print(local_sparse_matrix.shape)
if savename is not None:
pickle.dump(local_sparse_matrix, open('data/{}_local.pck'.format(savename), 'wb'))
# Now set up tourist array
tourist_users = set(tourist_connections.user_id)
tourist_businesses = set(tourist_connections.business_id)
tourist_data = tourist_connections['stars'].tolist()
tourist_col = tourist_connections.user_id.astype('category', categories=tourist_users).cat.codes
tourist_row = tourist_connections.business_id.astype('category', categories=tourist_businesses).cat.codes
assert(tourist_row[tourist_row<0].shape[0] == 0)
assert(tourist_col[tourist_col<0].shape[0] == 0)
tourist_sparse_matrix = csc_matrix((tourist_data, (tourist_row, tourist_col)),
shape=(len(tourist_businesses), len(tourist_users)))
print(tourist_sparse_matrix.shape)
if savename is not None:
pickle.dump(tourist_sparse_matrix, open('data/{}_tourist.pck'.format(savename), 'wb'))
# Verify that the businesses are indeed the same by checking that each business id is correctly tagged with the right row for an example row
test_ind = local_row.iloc[1] # 1 could be anything in the range of rows
biz_ind = local_row[local_row == test_ind].index
test_biz = local_connections.loc[biz_ind]
test_biz = set(test_biz['business_id'])
tour_ind = tourist_row[tourist_row == test_ind].index
tour_biz = tourist_connections.loc[tour_ind]
tour_biz = tour_biz['business_id']
matches = [tour_biz.iloc[i] in test_biz for i in range(tour_biz.shape[0])]
assert(all(matches))
return local_sparse_matrix, tourist_sparse_matrix
_, _ = get_city_array(cities[CITYNAME], connections, user_res, major_businesses, CITYNAME)
# In[135]:
# user_res.keys()
# LV_users = user_res[user_res['group'] == 1]
# LV_businesses = major_businesses[major_businesses['metro_area'] == 1]
# # LV_connections = connections[connections['user_id'].isin(m0_users['user_id'])]
# print(connections.shape)
# LV_connections = connections[connections['user_id'].isin(LV_users['user_id'])]
# print(LV_connections.shape)
# LV_connections = LV_connections[LV_connections['business_id'].isin(LV_businesses['business_id'])]
# print(LV_connections.shape)
# # connections.user_id
# LV_local = LV_connections.pivot(index='user_id', columns='business_id', values='stars')
# # print(LV_local)
# print(LV_local.shape)
# LV_nonnan = LV_local[np.isnan(LV_local) == False]
# # ~np.isnan(LV_local) == True
# print(LV_nonnan)
# # LV_sparse = scipy.sparse.csr_matrix(LV_local.values.T)
# # print(LV_sparse.shape)
# # c_maxes = connections.groupby(['user_id', 'business_id']).review_count.transform(max)
# # c2 = connections[connections.review_count == c_maxes]
# # # print(connections.shape)
# # # print(c2.shape)
# In[92]:
# LV_users = set(LV_connections.user_id)
# LV_businesses = set(LV_connections.business_id)
# data = LV_connections['stars'].tolist()
# col = LV_connections.user_id.astype('category', categories=LV_users).cat.codes
# row = LV_connections.business_id.astype('category', categories=LV_businesses).cat.codes
# sparse_matrix = csc_matrix((data, (row, col)), shape=(len(LV_businesses), len(LV_users)))