-
Notifications
You must be signed in to change notification settings - Fork 0
/
regression_solve.py
310 lines (294 loc) · 14.3 KB
/
regression_solve.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
#!/usr/bin/python
#_*_coding:UTF-8_*_
'''
author:jiaxin jiang
time:23:16
date:1/19/2019
function:using validation data decide train model to test test data
'''
import numpy as np
import pandas as pd
from sklearn.utils import shuffle
import os
from sklearn.model_selection import StratifiedKFold
import string
import keras
from keras.callbacks import ModelCheckpoint
from keras.callbacks import EarlyStopping
from keras.models import load_model
from keras.utils import multi_gpu_model
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import DataGenerator as dg
import get_modelv2_3
from keras.utils import plot_model
import sys
from keras import backend as K
import tensorflow as tf
from sklearn.linear_model import LinearRegression
import scipy
import seaborn as sns
# sns.set(color_codes=True)
import re
import warnings
warnings.filterwarnings('ignore')
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1,2,3' # n设置可见gpu7,3,4,0,1,4,23,4,5,66,7
class Regression_solve:
def __init__(self):
self.model=keras.Model()
# legend字体
self.font1 = {
'weight': 'normal',
'size': 16,
}
# 横纵坐标标题字体
self.font2 = {
'weight': 'normal',
'size': 23,
}
def change_model_2_regression(self):
'''
特比的,当把分类模型最后输出的激活函数单独拿出来当做一层之后,将分类模型修改成回归模型,只需要把最后一层去掉就可以了
'''
layer_name = self.model.layers[-2].name#倒数第二层
origin_outputs = self.model.get_layer(layer_name).output
inputs = self.model.input
# output = keras.layers.Dense(1, name='regression_output')(origin_outputs) # activation='relu',
self.model = keras.Model(inputs=inputs, outputs=origin_outputs)#outputs=output
print("generate regression model succeed!")
print(self.model.summary())
def mean_squared_error_l2(self, y_true, y_pred, lmbda=0.01):
cost = K.mean(K.square(y_pred - y_true))
# weights = self.model.get_weights()
weights = []
for layer in self.model.layers:
# print(layer)
weights = weights + layer.get_weights()
# print (weights)
result = tf.reduce_sum([tf.reduce_sum(tf.pow(wi, 2)) for wi in weights])
l2 = lmbda * result # K.sum([K.square(wi) for wi in weights])
return cost + l2
def predict_result(self,x_prot,x_comp,y_value,type):
predict_result = self.model.predict([x_prot, x_comp])
real = y_value
df = pd.DataFrame(predict_result, columns=['predicted'])
df['real'] = real
df['set'] = type
return df
def save_predict_result(self,x_prot,x_comp,y_value,model_name,type):
# 保存预测结
df=self.predict_result(x_prot,x_comp,y_value,type)
if not os.path.exists('predict_value'):
os.mkdir('predict_value')
df.to_csv('predict_value/regression_model_%s_%s_predict_result.csv' % (model_name,type), index=False)
def computer_parameter(self, df,type):
# 计算参数,画散点图
rmse = ((df['predicted'] - df['real']) ** 2).mean() ** 0.5
mae = (np.abs(df['predicted'] - df['real'])).mean()
corr = scipy.stats.pearsonr(df['predicted'], df['real'])
lr = LinearRegression()
lr.fit(df[['predicted']], df['real'])
y_ = lr.predict(df[['predicted']])
sd = (((df["real"] - y_) ** 2).sum() / (len(df) - 1)) ** 0.5
print("%10s set: RMSE=%.3f, MAE=%.3f, R=%.2f (p=%.2e), SD=%.3f" % (type, rmse, mae, *corr, sd))
return type, rmse, mae, corr, sd
def computer_parameter_draw_scatter_plot(self,x_prot, x_comp, y_value,model_name,type):
sns.set(context='paper', style='white')
sns.set_color_codes()
df = self.predict_result(x_prot, x_comp, y_value, type)
if all(df['real']>0):
xlimb_start=0
else:
xlimb_start=-10
if all(df['predicted']>0):
ylimb_start=0
else:
ylimb_start=-10
set_colors = {'train': 'b', 'validation': 'green', 'test': 'purple'}
grid = sns.jointplot('real', 'predicted', data=df, stat_func=None, color=set_colors[type],
space=0.0, size=4, ratio=4, s=20, edgecolor='w', ylim=(ylimb_start, 16),
xlim=(xlimb_start, 16)
)# s:点的大小;x-lim和y-lim
grid.ax_joint.set_xticks(range(xlimb_start, 16, 5))
grid.ax_joint.set_yticks(range(ylimb_start, 16, 5)) # 可单独画一张带负值的,但还是以真实正值为主
type, rmse, mae, corr, sd=self.computer_parameter(df,type)
grid.ax_joint.text(1, 14, type + ' set', fontsize=14) # 调整标题大小
grid.ax_joint.text(16, 19.5, 'RMSE: %.2f ' % rmse)
grid.ax_joint.text(16, 18.5, '(p): %.3f ' % corr[1])
grid.ax_joint.text(16, 17.5, 'R2: %.2f ' % corr[0])
grid.ax_joint.text(16, 16.5, 'SD: %.2f ' % sd)
grid.fig.savefig('%s_%s_scatter_plot.jpg' %(model_name,type), dpi=400)
def draw_loss_change(self,history,model_name):
history_dict = history.history
loss_values = history_dict['loss']
val_loss_values = history_dict['val_loss']
epochs = range(1, len(loss_values) + 1)
plt.figure(figsize=(10, 10))
plt.plot(epochs, loss_values, 'b', label='Training loss')
plt.plot(epochs, val_loss_values, 'r', label='Validation loss')
plt.title('Training and validation loss', self.font2)
plt.xlabel('Epochs', self.font2)
plt.ylabel('Loss', self.font2)
plt.legend(prop=self.font1)
plt.savefig('%s_regression_training_validation_loss.png'%model_name)
# plt.show()
##___________________________________________________________________-
plt.figure(figsize=(10, 10))
times = range(1, len(loss_values) + 1)
plt.plot(times[10:(len(loss_values) + 1)], loss_values[10:(len(loss_values) + 1)], 'b',
label='Training loss')
plt.plot(times[10:(len(loss_values) + 1)], val_loss_values[10:(len(loss_values) + 1)], 'r',
label='Validation loss')
plt.tick_params(labelsize=20)
plt.xlabel('Epochs', self.font2)
plt.ylabel('Loss', self.font2)
plt.legend(prop=self.font1)
plt.savefig('%s_regression_training_validation_loss2.png'%model_name)
def read_data(self, data_type, file,reinforced=False):
print("starting read %s data:" % data_type)
x_prot, x_comp, y = dg.multi_process_read_pro_com_file_regression(file,reinforced=reinforced)
print("%s data,%s, has been read succeed!" % (data_type, file))
print('x_prot.shape', x_prot.shape)
print('x_comp.shape', x_comp.shape)
print('y.shape', y.shape)
return x_prot, x_comp, y
def train_model(self,train_file,validation_file,model_name,lr = 0.0001,batch_size = 512):
# 一些设置项
alpha=0.3
epochs = 300 # 50
patience = 10
save_dir = os.path.join(os.getcwd(), 'saved_models')
print('1')
# 读训练数据
train_x_prot, train_x_comp, train_y=self.read_data("train",train_file,reinforced=False)#True
#读验证集
validation_x_prot, validation_x_comp, validation_y =self.read_data("validation",validation_file)
# 生成模型回归模型
self.model=get_modelv2_3.get_model9()#生成分类模型
self.change_model_2_regression()# 修改最后一层,使之成为回归问题
# regression model tain
log_filepath = './tmp/keras_log/Xavier_uniform/'
tb_cb = keras.callbacks.TensorBoard(log_dir=log_filepath, write_images=1, histogram_freq=1)
# 设置log的存储位置,将网络权值以图片格式保持在tensorboard中显示,设置每一个周期计算一次网络的
# 权值,每层输出值的分布直方图
optimizer = keras.optimizers.Adam(lr=lr)
self.model.compile(loss=self.mean_squared_error_l2, optimizer=optimizer,
metrics=['mse', 'mae'])#loss='mean_squared_error'
##早结束
early_stopping = EarlyStopping(monitor='val_loss', patience=patience)
## checkpoint
filepath = save_dir + "/loss-reduction-{epoch:02d}-{val_loss:.2f}-{val_mean_squared_error:.2f}.hdf5"
checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only='True',mode='min')
# 中途训练效果提升, 则将文件保存, 每提升一次, 保存一次,
# 训练模型
history = self.model.fit([train_x_prot, train_x_comp],
train_y,
shuffle=True,
batch_size=batch_size,
epochs=epochs,
validation_data=([validation_x_prot, validation_x_comp], validation_y),
callbacks=[early_stopping, checkpoint] # 回调函数,,tb_cb
)
# Save model and weights
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
model_path = os.path.join(save_dir, model_name + '_regression_train.h5')
self.model.save(model_path)
print('Saved trained model at %s ' % model_path)
# Score
score = self.model.evaluate([validation_x_prot, validation_x_comp], validation_y)
print(score)
# saving predict value
self.save_predict_result(train_x_prot, train_x_comp, train_y, model_name,'train')
self.save_predict_result(validation_x_prot, validation_x_comp,validation_y, model_name,'validation')
# train and validation loss change
self.draw_loss_change(history,model_name)
# computing parameters and drawing scatter plot
self.computer_parameter_draw_scatter_plot(train_x_prot, train_x_comp, train_y, model_name,'train')
self.computer_parameter_draw_scatter_plot(validation_x_prot, validation_x_comp, validation_y,model_name,
'validation')
def load_model_test(self,model_file,test_file,train_file=None,validation_file=None):
# read test data
x_prot_test, x_comp_test, y_test =self.read_data("test",test_file)
# load_model
self.model = load_model(model_file, custom_objects={'mean_squared_error_l2': self.mean_squared_error_l2})
# score on the data
print("scoring on the test data:")
result = self.model.evaluate([x_prot_test, x_comp_test], y_test)
print("loss, mse and mae:", result)
tmp= model_file.split('/')[-1]
model_name=re.findall(r"(.+?).hdf5", tmp)[0]
# saving predict value
self.save_predict_result(x_prot_test, x_comp_test, y_test, model_name,'test')
# computing parameters and drawing scatter plot
self.computer_parameter_draw_scatter_plot(x_prot_test, x_comp_test, y_test, model_name,'test')
#for train file
if train_file != None:
#read data
x_prot_train, x_comp_train, y_train = self.read_data("train", train_file)
#score
print("scoring on the train data:")
result = self.model.evaluate([x_prot_train, x_comp_train], y_train)
print("loss, mse and mae:", result)
# saving predict value
self.save_predict_result(x_prot_train, x_comp_train, y_train, model_name, 'train')
# computing parameters and drawing scatter plot
self.computer_parameter_draw_scatter_plot(x_prot_train, x_comp_train, y_train, model_name, 'train')
if validation_file != None:
# read data
x_prot_validation, x_comp_validation, y_validation = self.read_data("validation", validation_file)
# score
print("scoring on the validation data:")
result = self.model.evaluate([x_prot_validation, x_comp_validation], y_validation)
print("loss, mse and mae:", result)
# saving predict value
self.save_predict_result(x_prot_validation, x_comp_validation, y_validation, model_name, 'validation')
# computing parameters and drawing scatter plot
self.computer_parameter_draw_scatter_plot(x_prot_validation, x_comp_validation, y_validation,
model_name,'validation')
def load_model_predict(self,model_file,file):
# read data
x_prot, x_comp, y_label = dg.multi_process_read_pro_com_file(file)
# load_model
self.model = load_model(model_file, custom_objects={'mean_squared_error_l2': self.mean_squared_error_l2})
tmp= model_file.split('/')[-1]
model_name=re.findall(r"(.+?).hdf5", tmp)[0]
# saving predict value
self.save_predict_result(x_prot, x_comp, y_label, model_name,'none')
def main():
if len(sys.argv)==4:
#model train
train_file = sys.argv[1]
validation_file = sys.argv[2]
model_name = sys.argv[3]
print("train data is", train_file)
regression_model=Regression_solve()
regression_model.train_model(train_file,validation_file,model_name)
elif len(sys.argv)==3:
# model test
model_file = sys.argv[1]
test_file = sys.argv[2]
print("test data is", test_file)
regression_model = Regression_solve()
regression_model.load_model_test(model_file, test_file)
# regression_model.load_model_predict(model_file, test_file)
elif len(sys.argv)==5:
# model test
model_file = sys.argv[1]
test_file = sys.argv[2]
train_file = sys.argv[3]
validation_file = sys.argv[4]
regression_model = Regression_solve()
regression_model.load_model_test(model_file, test_file,train_file,validation_file)
else:
#
print("input parametes not illegal, please chec k and reinput!")
train_file ='../dataset/dataset_reg_train.txt'
validation_file = '../dataset/dataset_reg_vali.txt'
model_name = 'mode9_on_dataset_reg_4.25'
print("train data is", train_file)
regression_model = Regression_solve()
regression_model.train_model(train_file, validation_file, model_name)
if __name__ == '__main__':
main()