-
Notifications
You must be signed in to change notification settings - Fork 0
/
model_training.py
195 lines (177 loc) · 7.08 KB
/
model_training.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
import numpy as np
import pickle
import csv
from sklearn.ensemble import GradientBoostingRegressor, GradientBoostingClassifier
from sklearn.model_selection import KFold
from sklearn.metrics import r2_score, mean_absolute_error, accuracy_score
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
import warnings
warnings.filterwarnings("ignore")
def cross_validation_score(X_train, y_train, X_test, y_test):
pca = PCA(0.95)
pca.fit(X_train)
X_train = pca.transform(X_train)
X_test = pca.transform(X_test)
clf = GradientBoostingRegressor()
clf.fit(X_train, y_train)
prediction = clf.predict(X_test)
true_list = np.asarray(y_test)
prediction = np.asarray(prediction)
return np.mean(np.absolute(prediction.flatten()-true_list.flatten())/true_list.flatten())*100, pca.n_components_
def generate_missing(total_data_number, missing_data_rate, missing_feature_number):
missing_data_idx = np.random.choice(range(total_data_number), size=int(total_data_number*missing_data_rate), replace=False)
missing_feature_idx = []
for _ in range(int(total_data_number*missing_data_rate)):
missing_feature_idx.append(np.random.choice(range(5), size=missing_feature_number, replace=False))
return missing_data_idx, missing_feature_idx
def generate_missing_v2(total_data_number, missing_data_rate, missing_feature_number_list):
missing_data_idx = np.random.choice(range(total_data_number), size=int(total_data_number*missing_data_rate), replace=False)
missing_feature_idx = []
for _ in range(int(total_data_number*missing_data_rate)):
missing_feature_number = np.random.choice(missing_feature_number_list)
missing_feature_idx.append(np.random.choice(range(5), size=missing_feature_number, replace=False))
return missing_data_idx, missing_feature_idx
def fusion(X, Y, missing_data_idx, missing_feature_idx):
data = []
label = []
for i, dd in enumerate(X):
temp = np.zeros([1, dd.shape[1]])
count = 0.
if i in missing_data_idx:
idx = missing_data_idx.index(i)
for j, d in enumerate(dd):
if j not in missing_feature_idx[idx]:
count += 1
temp += d
else:
for j, d in enumerate(dd):
count += 1
temp += d
temp /= count
temp = temp.flatten()
data.append(temp)
label.append(Y[i])
return np.asarray(data), np.asarray(label)
def fusion_test(X, Y):
data = []
label = []
for i, dd in enumerate(X):
temp = np.zeros([1, dd.shape[1]])
count = 0.
for j, d in enumerate(dd):
count += 1
temp += d
temp /= count
temp = temp.flatten()
data.append(temp)
label.append(Y[i])
return np.asarray(data), np.asarray(label)
def remove_missing(X, Y, missing_data_idx, missing_feature_idx):
data = []
label = []
for i in range(len(X)):
if i not in missing_data_idx:
data.append(X[i].flatten())
label.append(Y[i])
return np.asarray(data), np.asarray(label)
def remove_missing_test(X, Y):
data = []
label = []
for i in range(len(X)):
data.append(X[i].flatten())
label.append(Y[i])
return np.asarray(data), np.asarray(label)
# for sphere function dataset
num_repeat = 10
with open('data.pkl', 'rb') as f:
data = pickle.load(f)
feature = data['feature']
label = data['label']
test_feature = feature[-200:]
test_label = label[-200:]
feature = feature[:-200]
label = label[:-200]
print (feature.shape, test_feature.shape)
missing_data_rate_list = np.asarray(range(1, 100))/100.
missing_feature_number_list = [1, 2, 3, 4]
feature_flatten = feature.reshape((len(feature), -1))
test_feature_flatten = test_feature.reshape((len(test_feature), -1))
ss = StandardScaler()
ss.fit(feature_flatten)
feature_flatten = ss.transform(feature_flatten)
test_feature_flatten = ss.transform(test_feature_flatten)
print (feature_flatten.shape, test_feature_flatten.shape)
logging = []
res, num = cross_validation_score(feature_flatten, label, test_feature_flatten, test_label)
a = ['nofusion', 0, 0, res, num, len(feature_flatten)]
print (a)
logging.append(a)
logging.append(['**', '**', '**', '**', '**', '**'])
for missing_data_rate in missing_data_rate_list:
for missing_feature_number in missing_feature_number_list:
temp = 0.
for _ in range(num_repeat):
missing_data_idx, missing_feature_idx = generate_missing(len(feature), missing_data_rate, missing_feature_number)
fused_data, fused_label = fusion(feature, label, list(missing_data_idx), missing_feature_idx)
test_fused_data, test_fused_label = fusion_test(test_feature, test_label)
ss = StandardScaler()
ss.fit(fused_data)
fused_data = ss.transform(fused_data)
test_fused_data = ss.transform(test_fused_data)
res, num = cross_validation_score(fused_data, fused_label, test_fused_data, test_fused_label)
temp += res
res = temp/num_repeat
a = ['fusion', missing_data_rate, missing_feature_number, res, num, len(fused_data)]
print (a)
logging.append(a)
temp = 0.
for _ in range(num_repeat):
remain_data, remain_label = remove_missing(feature, label, missing_data_idx, missing_feature_idx)
test_remain_data, test_remain_label = remove_missing_test(test_feature, test_label)
ss = StandardScaler()
ss.fit(remain_data)
remain_data = ss.transform(remain_data)
test_remain_data = ss.transform(test_remain_data)
res, num = cross_validation_score(remain_data, remain_label, test_remain_data, test_remain_label)
temp += res
res = temp/num_repeat
a = ['no fusion', missing_data_rate, missing_feature_number, res, num, len(remain_data)]
print (a)
logging.append(a)
logging.append(['**', '**', '**', '**', '**', '**'])
for missing_data_rate in missing_data_rate_list:
temp = 0.
for _ in range(num_repeat):
missing_data_idx, missing_feature_idx = generate_missing_v2(len(feature), missing_data_rate, missing_feature_number_list)
fused_data, fused_label = fusion(feature, label, list(missing_data_idx), missing_feature_idx)
test_fused_data, test_fused_label = fusion_test(test_feature, test_label)
ss = StandardScaler()
ss.fit(fused_data)
fused_data = ss.transform(fused_data)
test_fused_data = ss.transform(test_fused_data)
res, num = cross_validation_score(fused_data, fused_label, test_fused_data, test_fused_label)
temp += res
res = temp/num_repeat
a = ['fusion', missing_data_rate, missing_feature_number, res, num, len(fused_data)]
print (a)
logging.append(a)
temp = 0.
for _ in range(num_repeat):
remain_data, remain_label = remove_missing(feature, label, missing_data_idx, missing_feature_idx)
test_remain_data, test_remain_label = remove_missing_test(test_feature, test_label)
ss = StandardScaler()
ss.fit(remain_data)
remain_data = ss.transform(remain_data)
test_remain_data = ss.transform(test_remain_data)
res, num = cross_validation_score(remain_data, remain_label, test_remain_data, test_remain_label)
temp += res
res = temp/num_repeat
a = ['no fusion', missing_data_rate, missing_feature_number, res, num, len(remain_data)]
print (a)
logging.append(a)
with open('results.csv', 'w') as f:
writer = csv.writer(f, delimiter=',')
writer.writerow(['if fusion', 'missing_data_rate', 'missing_feature_number', 'MAE', '#_pca', '#_data'])
for i in range(len(logging)):
writer.writerow([logging[i][0], logging[i][1], logging[i][2], logging[i][3], logging[i][4], logging[i][5]])