-
Notifications
You must be signed in to change notification settings - Fork 99
/
Copy pathObtainPredict.py
134 lines (112 loc) · 5.55 KB
/
ObtainPredict.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
#!/usr/bin/python env
# -*- coding: utf-8 -*-
import numpy as np
from sklearn.linear_model import LogisticRegression
__author__ = 'Huang yi'
class PredictEmption(object):
def __init__(self):
pass
def DivideSet(self):
# 第一次加载重写txt文件到numpy的格式npy中。函数依次运行一次。
train_data = np.loadtxt('data_features.txt')
train_data_label = np.loadtxt('data_labels.txt')
np.save('data_features.npy', train_data)
np.save('data_labels.npy', train_data_label)
fop1 = open('feature_pos.txt', 'w+')
fop2 = open('feature_neg.txt', 'w+')
train_data = np.load('data_features.npy')
train_data_label = np.load('data_labels.npy')
label_size = train_data_label.size
# 转置
train_data_label.shape = (1, label_size)
train_data_label_transpose = np.transpose(train_data_label)
# 将label加到每行特征的最后
line = data_with_label = np.hstack((train_data, train_data_label_transpose))
# 把正例和负例分着放入feature_pos.txt和feature_neg.txt中
for i in range(0, label_size):
if(line[i][-1]):
fop1.write('%d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d' %
(line[i][0],line[i][1],line[i][2],line[i][3],line[i][4],line[i][5],line[i][6],line[i][7],line[i][8],line[i][9],
line[i][10],line[i][11],line[i][12],line[i][13],line[i][14],line[i][15],line[i][16],line[i][17],line[i][18],line[i][19]))
fop1.write('\n')
else:
fop2.write('%d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d %d' %
(line[i][0],line[i][1],line[i][2],line[i][3],line[i][4],line[i][5],line[i][6],line[i][7],line[i][8],line[i][9],
line[i][10],line[i][11],line[i][12],line[i][13],line[i][14],line[i][15],line[i][16],line[i][17],line[i][18],line[i][19]))
fop2.write('\n')
fop1.close()
fop2.close()
feature_pos = np.loadtxt('feature_pos.txt')
feature_neg = np.loadtxt('feature_neg.txt')
np.save('feature_p.npy', feature_pos)
np.save('feature_n.npy', feature_neg)
def GenTrainTestSet(self):
feature_pos = np.load('feature_p.npy')
feature_neg = np.load('feature_n.npy')
# 打乱生成随机序列
np.random.shuffle(feature_pos)
np.random.shuffle(feature_neg)
TrainSet = np.vstack((feature_pos[0:800], feature_neg[0:500000]))
TestSet = np.vstack((feature_pos[800:999], feature_neg[500000:750000]))
np.save('TrainSet.npy', TrainSet)
np.save('TestSet.npy', TestSet)
def TrainAndPredict(self, train_data, train_data_label, predict_data):
# 应用Logistic回归进行预测
LR = LogisticRegression()
LR.fit(train_data, train_data_label)
predict_labels = LR.predict(predict_data)
# 0为不买的概率，1为买的概率
predict_proba = LR.predict_proba(predict_data)[:, -1]
return predict_labels, predict_proba
def GetF1(self, predict_list, real_list):
_list = [x for x in predict_list if x in real_list]
equal_num = len(_list)
prediction_set_num = len(predict_list)
reference_set_num = len(real_list)
precision = float(equal_num) / prediction_set_num
recall = float(equal_num) / reference_set_num
f1 = float(2 * precision * recall) / (precision + recall)
return precision, recall, f1
# 用30day的TrainSet和TestSet测试，选出最优阈值
def TestPredict(self):
train_set = np.load('TrainSet.npy')
test_set = np.load('TestSet.npy')
train_data = train_set[:, 0:-1] # 去除label后的特征数据
train_data_label = train_set[:, -1] # label
test_data = test_set[:, 0:-1] # 不需要label
test_data_label = test_set[:, -1] # label
predict_labels, predict_proba = self.TrainAndPredict(train_data, train_data_label, test_data)
# 测试出最优的阈值设置
# best_th = 0.014
for i in range(0, 2):
rough_set = float(i) / 10
for j in range(0, 100):
tiny_set = float(j)/1000
th = rough_set + tiny_set
# 预测出的购买的list
index = np.array(range(0, len(predict_proba)))
# 大于阈值对应的索引
index_predict = index[predict_proba > th]
index_predict = list(index_predict)
# 真实购买的list
index_real = index[test_data_label > 0]
precision, recall, f1 = self.GetF1(index_predict, index_real)
print 'TH %s : %s %s %s | %s %s' % (th, precision, recall, f1, len(index_predict), len(index_real))
# index = np.array(range(0, len(predict_proba)))
# # 大于阈值对应的索引
# index_predict = index[predict_proba > best_th]
# index_predict = list(index_predict)
# # 真实购买的list
# index_real = index[test_data_label > 0]
# precision, recall, f1 = self.GetF1(index_predict, index_real)
# print 'TH %s : %s %s %s' % (best_th, precision, recall, f1)
if __name__ == '__main__':
# predict_data = np.load('predict_data.txt')
# PE.TrainAndPredict(train_data, train_data_label, predict_data)
# LR = LogisticRegression()
# ret = LR.fit(train_data, train_data_label)
# predict_labels = LR.predict(train_data[3000:3005])
PE = PredictEmption()
# PE.DivideSet()
# PE.GenTrainTestSet()
PE.TestPredict()