-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathevaluation.py
167 lines (116 loc) · 7.57 KB
/
evaluation.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
# -*- coding: utf-8 -*-
import torch
import numpy as np
class Evaluation:
def __init__(self, dataset, dataloader, user_count, h0_strategy, trainer, setting):
self.dataset = dataset
self.dataloader = dataloader
self.user_count = user_count
self.h0_strategy = h0_strategy
self.trainer = trainer
self.setting = setting
def evaluate(self):
h = self.h0_strategy.on_init(self.setting.batch_size, self.setting.device)
# (number_of_layer:1, batch_size:256, hidden_size:16)
with torch.no_grad():
iter_cnt = 0
recall1_poi = 0
recall5_poi = 0
recall10_poi = 0
average_precision_poi = 0
recall1_catg = 0
recall5_catg = 0
recall10_catg = 0
average_precision_catg = 0
recall1_catgLayer = 0
recall5_catgLayer = 0
recall10_catgLayer = 0
average_precision_catgLayer = 0
for i,(x_user, x_tf, x_tb, x_tsf, x_tsb, x_cof, x_cob, x_poi_f, x_poi_b, x_catg_f, x_catg_b, x_catgLayer_f, x_catgLayer_b, y_tsecond, y_tslot, y_coord, y_poi, y_catg, y_catgLayer) in enumerate(self.dataloader):
length = len(x_user)
h = h[:,:length,:]
iter_cnt = iter_cnt + length
#
x_user = x_user.squeeze().to(self.setting.device)
x_tf = torch.transpose(x_tf.squeeze(),0,1).to(self.setting.device)
x_tb = torch.transpose(x_tb.squeeze(),0,1).to(self.setting.device)
x_tsf = torch.transpose(x_tsf.squeeze(),0,1).to(self.setting.device)
x_tsb = torch.transpose(x_tsb.squeeze(),0,1).to(self.setting.device)
x_cof = torch.transpose(x_cof.squeeze(),0,1).to(self.setting.device)
x_cob = torch.transpose(x_cob.squeeze(),0,1).to(self.setting.device)
x_poi_f = torch.transpose(x_poi_f.squeeze(),0,1).to(self.setting.device)
x_poi_b = torch.transpose(x_poi_b.squeeze(),0,1).to(self.setting.device)
x_catg_f = torch.transpose(x_catg_f.squeeze(),0,1).to(self.setting.device)
x_catg_b = torch.transpose(x_catg_b.squeeze(),0,1).to(self.setting.device)
x_catgLayer_f = torch.transpose(x_catgLayer_f.squeeze(),0,1).to(self.setting.device)
x_catgLayer_b = torch.transpose(x_catgLayer_b.squeeze(),0,1).to(self.setting.device)
y_tsecond = y_tsecond.squeeze().to(self.setting.device)
y_tslot = y_tslot.squeeze().to(self.setting.device)
y_coord = y_coord.squeeze().to(self.setting.device)
y_poi = y_poi.squeeze()
y_catg = y_catg.squeeze()
y_catgLayer = y_catgLayer.squeeze()
y_pred_poi, y_pred_catgLayer = self.trainer.evaluate(h, x_user, x_tf, x_tb, x_tsf, x_tsb, x_cof, x_cob, x_poi_f, x_poi_b, x_catg_f, x_catg_b, x_catgLayer_f, x_catgLayer_b, y_tsecond, y_tslot, y_coord, y_poi, y_catg, y_catgLayer)
# (batchsize, loc_count)和（batchsize，catg_loc）
for j in range(length):
'''for POI'''
o_poi = y_pred_poi[j]
# o_poi = o_poi.cpu().detach().numpy()
o_poi = o_poi.cpu().numpy()
ind_poi = np.argpartition(o_poi, -10)[-10:]
r_poi = ind_poi[np.argsort(-o_poi[ind_poi], axis=0)]
r_poi = torch.tensor(r_poi)
# compute for MAP
y_poi_index = y_poi[j] # ground truth_index
y_poi_value = o_poi[y_poi_index]
upper_poi = np.where(o_poi > y_poi_value)[0]
map_poi = 1. / (1+len(upper_poi))
recall1_poi += y_poi_index in r_poi[:1]
recall5_poi += y_poi_index in r_poi[:5]
recall10_poi+= y_poi_index in r_poi[:10]
average_precision_poi += map_poi
# '''For category'''
# o_catg = y_pred_catg[j]
# # o_catg = o_catg.cpu().detach().numpy()
# o_catg = o_catg.cpu().numpy()
# ind_catg = np.argpartition(o_catg, -10)[-10:]
#
# r_catg = ind_catg[np.argsort(-o_catg[ind_catg], axis=0)]
# r_catg = torch.tensor(r_catg)
#
# y_catg_index = y_catg[j]
# y_catg_value = o_catg[y_catg_index]
# upper_catg = np.where(o_catg > y_catg_value)[0]
# map_catg = 1. / (1 + len(upper_catg))
#
# recall1_catg += y_catg_index in r_catg[:1]
# recall5_catg += y_catg_index in r_catg[:5]
# recall10_catg += y_catg_index in r_catg[:10]
# average_precision_catg += map_catg
'''For categoryLayer'''
o_catgLayer = y_pred_catgLayer[j]
# o_catg = o_catg.cpu().detach().numpy()
o_catgLayer = o_catgLayer.cpu().numpy()
ind_catgLayer = np.argpartition(o_catgLayer, -10)[-10:]
r_catgLayer = ind_catgLayer[np.argsort(-o_catgLayer[ind_catgLayer], axis=0)]
r_catgLayer = torch.tensor(r_catgLayer)
y_catgLayer_index = y_catgLayer[j]
y_catgLayer_value = o_catgLayer[y_catgLayer_index]
upper_catgLayer = np.where(o_catgLayer > y_catgLayer_value)[0]
map_catgLayer = 1./(1+len(upper_catgLayer))
recall1_catgLayer += y_catgLayer_index in r_catgLayer[:1]
recall5_catgLayer += y_catgLayer_index in r_catgLayer[:5]
recall10_catgLayer += y_catgLayer_index in r_catgLayer[:10]
average_precision_catgLayer += map_catgLayer
print('POI:Recall1:', recall1_poi/iter_cnt)
print('POI:Recall5:', recall5_poi/iter_cnt)
print('POI:Recall10:', recall10_poi/iter_cnt)
print('POI:MAP:', average_precision_poi/iter_cnt)
# print('Catg:Recall1:', recall1_catg/iter_cnt)
# print('Catg:Recall5:', recall5_catg/iter_cnt)
# print('Catg:Recall10:', recall10_catg/iter_cnt)
# print('Catg:MAP:', average_precision_catg/iter_cnt)
print('CatgLayer:Recall1:', recall1_catgLayer / iter_cnt)
print('CatgLayer:Recall5:', recall5_catgLayer / iter_cnt)
print('CatgLayer:Recall10:', recall10_catgLayer / iter_cnt)
print('CatgLayer:MAP:', average_precision_catgLayer / iter_cnt)