|
@@ -9,11 +9,9 @@ from imblearn.over_sampling import RandomOverSampler
|
9
|
9
|
|
10
|
10
|
|
11
|
11
|
def read_data(path):
|
12
|
|
- train_x=[]
|
13
|
|
- train_y=[]
|
14
|
12
|
lines = []
|
15
|
13
|
with open(path) as f:
|
16
|
|
- for x in range(20000):
|
|
14
|
+ for x in range(40000):
|
17
|
15
|
lines.append(eval(f.readline().strip()))
|
18
|
16
|
|
19
|
17
|
random.shuffle(lines)
|
|
@@ -36,47 +34,38 @@ def read_data(path):
|
36
|
34
|
|
37
|
35
|
return X_resampled,y_resampled,np.array(test_x),np.array(test_y)
|
38
|
36
|
|
39
|
|
-train_x,train_y,test_x,test_y=read_data("D:\\data\\quantization\\stock4.log")
|
40
|
|
-
|
41
|
|
-model = Sequential()
|
42
|
|
-model.add(Dense(units=425, input_dim=166, activation='relu'))
|
43
|
|
-model.add(Dense(units=325, activation='relu'))
|
44
|
|
-model.add(Dense(units=325, activation='relu'))
|
45
|
|
-model.add(Dropout(0.2))
|
46
|
|
-# model.add(Dense(units=325, activation='relu'))
|
47
|
|
-# model.add(Dropout(0.2))
|
48
|
|
-# model.add(Dense(units=325, activation='relu'))
|
49
|
|
-# model.add(Dropout(0.2))
|
50
|
|
-# model.add(Dense(units=325, activation='relu'))
|
51
|
|
-# model.add(Dropout(0.2))
|
52
|
|
-# model.add(Dense(units=325, activation='relu'))
|
53
|
|
-# model.add(Dropout(0.2))
|
54
|
|
-# model.add(Dense(units=225, activation='relu'))
|
55
|
|
-# model.add(Dense(units=225, activation='relu'))
|
56
|
|
-# model.add(Dense(units=225, activation='relu'))
|
57
|
|
-model.add(Dense(units=225, activation='relu'))
|
58
|
|
-model.add(Dropout(0.2))
|
59
|
|
-# model.add(Dense(units=225, activation='relu'))
|
60
|
|
-model.add(Dense(units=125, activation='relu'))
|
61
|
|
-model.add(Dropout(0.2))
|
62
|
|
-# model.add(Dense(units=125, activation='relu'))
|
63
|
|
-model.add(Dense(units=166, activation='relu'))
|
64
|
|
-model.add(Dense(units=8, activation='softmax'))
|
65
|
|
-model.compile(loss='categorical_crossentropy', optimizer="adam",metrics=['accuracy'])
|
66
|
|
-
|
67
|
|
-print("Starting training ")
|
68
|
|
-h=model.fit(train_x, train_y, batch_size=64, epochs=26, shuffle=True)
|
69
|
|
-score = model.evaluate(test_x, test_y)
|
70
|
|
-print(score)
|
71
|
|
-print('Test score:', score[0])
|
72
|
|
-print('Test accuracy:', score[1])
|
73
|
|
-
|
74
|
|
-path="model_seq.h5"
|
75
|
|
-model.save(path)
|
76
|
|
-model=None
|
77
|
|
-
|
78
|
|
-
|
79
|
|
-model=load_model(path)
|
80
|
|
-result=model.predict(test_x)
|
81
|
|
-print(result)
|
82
|
|
-print(test_y)
|
|
37
|
+
|
|
38
|
+def train(input_dim=400, result_class=3, file_path="D:\\data\\quantization\\stock6.log", model_name=''):
|
|
39
|
+ train_x,train_y,test_x,test_y=read_data(file_path)
|
|
40
|
+
|
|
41
|
+ model = Sequential()
|
|
42
|
+ model.add(Dense(units=625, input_dim=input_dim, activation='relu'))
|
|
43
|
+ model.add(Dense(units=525, activation='relu'))
|
|
44
|
+ model.add(Dense(units=525, activation='relu'))
|
|
45
|
+ model.add(Dropout(0.2))
|
|
46
|
+ model.add(Dense(units=525, activation='relu'))
|
|
47
|
+ model.add(Dropout(0.2))
|
|
48
|
+ model.add(Dense(units=325, activation='relu'))
|
|
49
|
+ model.add(Dropout(0.2))
|
|
50
|
+ model.add(Dense(units=266, activation='relu'))
|
|
51
|
+ model.add(Dense(units=result_class, activation='softmax'))
|
|
52
|
+ model.compile(loss='categorical_crossentropy', optimizer="adam",metrics=['accuracy'])
|
|
53
|
+
|
|
54
|
+ print("Starting training ")
|
|
55
|
+ h=model.fit(train_x, train_y, batch_size=64, epochs=26, shuffle=True)
|
|
56
|
+ score = model.evaluate(test_x, test_y)
|
|
57
|
+ print(score)
|
|
58
|
+ print('Test score:', score[0])
|
|
59
|
+ print('Test accuracy:', score[1])
|
|
60
|
+
|
|
61
|
+ model.save(model_name)
|
|
62
|
+
|
|
63
|
+ model=None
|
|
64
|
+ model=load_model(model_name)
|
|
65
|
+ result=model.predict(test_x)
|
|
66
|
+ print(result)
|
|
67
|
+ print(test_y)
|
|
68
|
+
|
|
69
|
+
|
|
70
|
+if __name__ == '__main__':
|
|
71
|
+ train(input_dim=46, result_class=5, file_path="D:\\data\\quantization\\stock6_5.log", model_name='5d_dnn_seq.h5')
|