-
Notifications
You must be signed in to change notification settings - Fork 1
/
algo2dl2.py
208 lines (142 loc) · 5.08 KB
/
algo2dl2.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Wed Apr 5 13:06:13 2017
@author: xlychee
"""
import pandas as pd
import time
import random
import numpy as np
import datetime
import tensorflow as tf
#import matplotlib.pyplot as plt
import math
import pickle
from sklearn.model_selection import train_test_split
from datetime import timedelta
def curtime():
return time.asctime(time.localtime(time.time()))
print curtime()+" Program begin "
df = pd.read_csv('../train_new_41')
df,dftest = train_test_split(df, test_size = 0.2)
sess = tf.InteractiveSession()
places = np.array(list(set(df['last_place'].unique()) |set(df['now_place'].unique())))
num_places = len(places)
map_places = {}
counter = 0
for p in places:
if p not in map_places:
map_places[p] = counter
counter +=1
users = df['user_id'].unique()
num_users = len(users)
map_users = {}
counter = 0
for u in users:
if u not in map_users:
map_users[u]=counter
counter+=1
times = df['now_time'].unique()
num_times = len(times)
map_times ={}
counter = 0
for t in times:
if t not in map_times:
map_times[t] = counter
counter+=1
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
def bias_variable(shape):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
#dim_features = 100
w_p = weight_variable([num_places,100])
b_p = bias_variable([100])
#x_p = tf.placeholder(tf.float32, shape=[None, num_places])
x_p = tf.placeholder(tf.int32)
x_p_1 = tf.one_hot(x_p,depth = num_places)
x_p_2 = tf.reshape(x_p_1, [-1, num_places])
feature_p = tf.matmul(x_p_2, w_p) +b_p
w_u = weight_variable([num_users, 100])
b_u = bias_variable([100])
x_u = tf.placeholder(tf.float32, shape=[None, num_users])
feature_u = tf.matmul(x_u, w_u) +b_u
w_t = weight_variable([num_times, 30])
b_t = bias_variable([30])
x_t = tf.placeholder(tf.float32, shape=[None, num_times])
feature_t = tf.matmul(x_t, w_t) +b_t
w1 = weight_variable([100,200])
b1 = bias_variable([200])
layer1 = tf.matmul(feature_p,w1)+b1
w2 = weight_variable([100,200])
b2 = bias_variable([200])
layer1 += tf.matmul(feature_u,w2)+b2
w3 = weight_variable([30,200])
b3 = bias_variable([200])
layer1 += tf.matmul(feature_t,w3)+b3
layer1 = tf.nn.relu(layer1)
ww = weight_variable([200,num_places])
bb = bias_variable([num_places])
output = tf.matmul(layer1,ww) + bb
y_ = tf.placeholder(tf.float32, shape=[None, num_places])
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=output))
regularizers = tf.nn.l2_loss(w_p)+ tf.nn.l2_loss(w_u)\
+ tf.nn.l2_loss(w_t) + tf.nn.l2_loss(w1) \
+ tf.nn.l2_loss(w2) + tf.nn.l2_loss(w3) \
+ tf.nn.l2_loss(ww)
loss = cross_entropy + regularizers
train_step = tf.train.AdamOptimizer(1e-4).minimize(loss)
sess.run(tf.global_variables_initializer())
def predict(u_id,lq_id,t_id):
x_u_batch = tf.one_hot([u_id],depth = num_users).eval()
x_t_batch = tf.one_hot([t_id],depth = num_times).eval()
feed_val = {x_p: lq_id, x_u: x_u_batch, x_t: x_t_batch}
result_ndarray = output.eval(feed_dict=feed_val)
top1 = result_ndarray.argmax()
result_ndarray[0][top1] = -float('Inf')
top2 = result_ndarray.argmax()
result_ndarray[0][top2] = -float('Inf')
top3 = result_ndarray.argmax()
result_ndarray[0][top3] = -float('Inf')
top4 = result_ndarray.argmax()
result_ndarray[0][top4] = -float('Inf')
top5 = result_ndarray.argmax()
recommend = [top1,top2,top3,top4,top5]
return recommend
print curtime()+" Training begin "
begintime = time.time()
total_steps = 50000
hit = 0
for i in range(total_steps):
u_id = map_users[df.iloc[i]['user_id']]
lq_id = map_places[df.iloc[i]['last_place']]
lp_id = map_places[df.iloc[i]['now_place']]
t_id = map_times[df.iloc[i]['now_time']]
x_u_batch = tf.one_hot([u_id],depth = num_users).eval()
x_p_batch = tf.one_hot([lq_id],depth = num_places).eval()
x_t_batch = tf.one_hot([t_id],depth = num_times).eval()
y_batch = tf.one_hot([lp_id],depth = num_places).eval()
feed = {x_p: lq_id, x_u: x_u_batch, x_t: x_t_batch,y_: y_batch}
if i%10 == 0 and i!=0:
used_time = str(timedelta(seconds=int(time.time()-begintime)))
#print "here", u_id, lq_id, lp_id, t_id
current_accuracy = float(hit)/i
print("step %d/%d, used time %s, current train accuracy %g" %(i,total_steps, used_time, current_accuracy))
print 'loss',loss_val
train_step.run(feed_dict = feed)
loss_val = loss.eval(feed_dict = feed)
if lp_id in predict(u_id,lq_id,t_id):
hit += 1
test_num = 0
accuracy = 0
for i in range(test_num):
u_id = map_users[dftest.iloc[i]['user_id']]
lq_id = map_places[dftest.iloc[i]['last_place']]
lp_id = map_places[dftest.iloc[i]['now_place']]
t_id = map_times[dftest.iloc[i]['now_time']]
if lp_id in predict(u_id,lq_id,t_id):
hit += 1
print 'current test accuracy',float(hit)/test_num