-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDocument_internal_data_enhancement.py
316 lines (284 loc) · 12.1 KB
/
Document_internal_data_enhancement.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
# coding=utf-8
import os.path
from __init__ import SS3
from util import Dataset
import random as ra
from sklearn.metrics import f1_score,confusion_matrix
from sklearn import metrics
import math
import GF_SCORE
import pickle
class DAEB:
clf = SS3()
GF = []
prediction = []
__strength__ = []
def gmean(matrixes):
'''gmean函数'''
m = matrixes
Recall = m[1][1] / (m[1][1] + m[1][0])
Specificity = m[0][0] / (m[0][1] + m[0][0])
GMean_Score = math.sqrt(Recall * Specificity)
# print("G-Mean Score: "+str(GMean_Score))
return GMean_Score
def build_data(x, y, n, strength: int):
'''
:param x: x_data
:param y: y_data
:param n: 划分比例
:param strength: 文档内增强系数
:return: x_train,y_train,x_test,y_test
'''
len_data = len(y)
len_po = 0
len_ne = 0
for i in range(0, int(len_data)):
if y[i] == 'po':
len_po = len_po + 1
if y[i] == 'ne':
len_ne = len_ne + 1
x_test = []
y_test = []
x_train = []
y_train = []
'''ne'''
arr_ne = ra.sample(range(0, len_ne), len_ne)
for i in range(0, int(len_ne / n)):
x_test.append(x[arr_ne[i]])
y_test.append(y[arr_ne[i]])
for i in range(0, int((n - 1) * len_ne / n)):
a = x[arr_ne[int(len_ne / n) + i]]
x_train.append(strength * a[500:-100] + x[arr_ne[int(len_ne / n) + i]])
y_train.append(y[arr_ne[int(len_ne / n) + i]])
'''po'''
arr_po = ra.sample(range(len_ne, len_ne + len_po), len_po)
for j in range(0, int(len_po / n)):
x_test.append(x[arr_po[j]])
y_test.append(y[arr_po[j]])
for j in range(0, int((int((n - 1) * len_po / n)) / 1)):
x_train.append(x[arr_po[int(len_po / n) + j]])
y_train.append(y[arr_po[int(len_po / n) + j]])
return x_train, y_train, x_test, y_test
def random_sample(X_train, Y_train, rate):
"""
分层随机抽样:(保证少量数据集先进行数据载入）
:param X_train: 待采样样本
:param Y_train: 采样个数
:param rate:采样比例
"""
x_train = []
y_train = []
'''计算采样规模'''
lenth = len(Y_train)
lenth_ne = 0
lenth_po = 0
for i in range(0, lenth):
if Y_train[i] == 'ne':
lenth_ne = lenth_ne + 1
else:
lenth_po = lenth_po + 1
'''ne层采样'''
for i in range(0, int(lenth_ne * rate)): # 设置采样次数
random = ra.randint(1, lenth_ne - 1)
x_train.append(X_train[random])
y_train.append(Y_train[random])
'''po层采样'''
for j in range(0, int(lenth_po * rate)):
random2 = ra.randint(lenth_ne, lenth_po - 1)
x_train.append(X_train[random2])
y_train.append(Y_train[random2])
return x_train, y_train
def bagging_train(X_train, Y_train, x_test, y_test, rate, threshold):
'''
:param X_train: 训练集
:param Y_train: 标签
:param x_test: 测试集
:param y_test: 标签
:param rate: 采样频率
:param threshold: 丢弃阈值
:return
'''
'''训练第i个模型'''
Invalid_model = 0
for i in range(0, 5):
'''5个模型集成'''
x_train, y_train = DAEB.random_sample(X_train, Y_train, rate)
DAEB.clf.set_hyperparameters(s=0.32, l=1.62, p=1.1)
DAEB.clf.train(x_train, y_train, n_grams=i + 2)
y_pred = DAEB.clf.predict(1, x_test)
'''保存预测结果'''
gf_score = GF_SCORE.gf_score(y_test, y_pred, 1, 1)
if gf_score > threshold:
if i == 0:
pickle.dump(DAEB.clf, file=open('./model1.pkl', 'wb+'))
if i == 1:
pickle.dump(DAEB.clf, file=open('./model2.pkl', 'wb+'))
if i == 2:
pickle.dump(DAEB.clf, file=open('./model3.pkl', 'wb+'))
if i == 3:
pickle.dump(DAEB.clf, file=open('./model4.pkl', 'wb+'))
if i == 4:
pickle.dump(DAEB.clf, file=open('./model5.pkl', 'wb+'))
else:
Invalid_model = Invalid_model + 1
if Invalid_model == 5:
quit('丢弃阈值过高，无可用模型，请重新设置！')
print('训练完成')
def bagging_predict(x_test):
'''训练第i个模型'''
y_pred = []
live_learner = 0
for i in range(0, 5):
'''5个模型集成'''
if i == 0 and os.path.isfile('model1.pkl'):
DAEB.clf = pickle.load(file=open('./model1.pkl', 'rb'))
os.remove('./model1.pkl')
live_learner = live_learner + 1
if i == 1 and os.path.isfile('model2.pkl'):
DAEB.clf = pickle.load(file=open('./model2.pkl', 'rb'))
os.remove('./model2.pkl')
live_learner = live_learner + 1
if i == 2 and os.path.isfile('model3.pkl'):
DAEB.clf = pickle.load(file=open('./model3.pkl', 'rb'))
os.remove('./model3.pkl')
live_learner = live_learner + 1
if i == 3 and os.path.isfile('model4.pkl'):
DAEB.clf = pickle.load(file=open('./model4.pkl', 'rb'))
os.remove('./model4.pkl')
live_learner = live_learner + 1
if i == 4 and os.path.isfile('model5.pkl'):
DAEB.clf = pickle.load(file=open('./model5.pkl', 'rb'))
os.remove('./model5.pkl')
live_learner = live_learner + 1
DAEB.clf.set_hyperparameters(s=0.32, l=1.62, p=1.1)
y_pred.append(DAEB.clf.predict(1, x_test))
'''投票得出结果'''
if live_learner == 1:
return y_pred[0]
Y_pred = []
test_len = len(y_pred[0])
if live_learner > 1:
for i in range(0, test_len):
live_flag = live_learner
votes_number = 1
if live_flag > 1:
if y_pred[0][i] == y_pred[1][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[1][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[2][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[2][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[3][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[3][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[4][i]:
votes_number = votes_number + 1
else:
different = y_pred[4][i]
if votes_number > (live_learner / 2): # 投票数大于存活学习器的一半
Y_pred.append(y_pred[0][i])
else:
Y_pred.append(different)
return Y_pred
def predict(x_test):
'''训练第i个模型'''
y_pred = []
live_learner = 0
for i in range(0, 5):
'''5个模型集成'''
if i == 0 and os.path.isfile('model1.pkl'):
DAEB.clf = pickle.load(file=open('./model1.pkl', 'rb'))
live_learner = live_learner + 1
if i == 1 and os.path.isfile('model2.pkl'):
DAEB.clf = pickle.load(file=open('./model2.pkl', 'rb'))
live_learner = live_learner + 1
if i == 2 and os.path.isfile('model3.pkl'):
DAEB.clf = pickle.load(file=open('./model3.pkl', 'rb'))
live_learner = live_learner + 1
if i == 3 and os.path.isfile('model4.pkl'):
DAEB.clf = pickle.load(file=open('./model4.pkl', 'rb'))
live_learner = live_learner + 1
if i == 4 and os.path.isfile('model5.pkl'):
DAEB.clf = pickle.load(file=open('./model5.pkl', 'rb'))
live_learner = live_learner + 1
DAEB.clf.set_hyperparameters(s=0.32, l=1.62, p=1.1)
y_pred.append(DAEB.clf.predict(1, x_test))
'''投票得出结果'''
if live_learner == 1:
return y_pred[0]
Y_pred = []
test_len = len(y_pred[0])
if live_learner > 1:
for i in range(0, test_len):
live_flag = live_learner
votes_number = 1
if live_flag > 1:
if y_pred[0][i] == y_pred[1][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[1][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[2][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[2][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[3][i]:
votes_number = votes_number + 1
live_flag = live_flag - 1
else:
different = y_pred[3][i]
live_flag = live_flag - 1
if live_flag > 1:
if y_pred[0][i] == y_pred[4][i]:
votes_number = votes_number + 1
else:
different = y_pred[4][i]
if votes_number > (live_learner / 2): # 投票数大于存活学习器的一半
Y_pred.append(y_pred[0][i])
else:
Y_pred.append(different)
return Y_pred
def DAEB_bagging_fit(X_DATA, Y_DATA, rate, down: int, up: int, threshold):
"""
分层增强随机抽样:(保证少量数据集先进行数据载入）
:param X_DATA,Y_DATA: 待采样样本
:param rate: 采样频率
down:搜索下界
up:搜索上界
return 最优预测结果
"""
for i in range(down, up):
X_train, Y_train, x_test, y_test = DAEB.build_data(X_DATA, Y_DATA, 5, i)
DAEB.bagging_train(X_train, Y_train, x_test, y_test, rate, threshold)
y_pred = DAEB.bagging_predict(x_test)
# GF
GFScore = GF_SCORE.gf_score(y_test, y_pred, 1, 1)
DAEB.GF.append(GFScore)
DAEB.prediction.append(y_pred)
DAEB.__strength__.append(i)
index = DAEB.GF.index(max(DAEB.GF))
best = DAEB.__strength__[index]
pre = DAEB.prediction[index]
print("最优采样系数: ", best)
X_train, Y_train, x_test, y_test = DAEB.build_data(X_DATA, Y_DATA, 5, best)
DAEB.bagging_train(X_train, Y_train, x_test, y_test, rate, threshold)
return pre
#X_DATA , Y_DATA = Dataset.load_from_files("C:/Users/GYM/Desktop/大创与互联网+：基于集成学习的心理状态分析与评测系统/代码/ss3 and t-ss3/erisk2018/eRisk2018/2018/task 1 - depression (test split, train split is 2017 data)/kflod")
#pre,y_test=DAEB_bagging_fit(X_DATA,Y_DATA,2/3,1,3,0.6)