回答編集履歴
1
edit
test
CHANGED
|
@@ -11,3 +11,125 @@
|
|
|
11
11
|
|
|
12
12
|
|
|
13
13
|
書き写したのですか？
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
|
|
17
|
+
---
|
|
18
|
+
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
とりあえず動くものを載せておくので参考にしてみてください。
|
|
22
|
+
|
|
23
|
+
```python
|
|
24
|
+
|
|
25
|
+
import pandas as pd
|
|
26
|
+
|
|
27
|
+
import numpy as np
|
|
28
|
+
|
|
29
|
+
import keras
|
|
30
|
+
|
|
31
|
+
from keras.models import Sequential
|
|
32
|
+
|
|
33
|
+
from keras.layers import LSTM, Dense, Activation, Dropout
|
|
34
|
+
|
|
35
|
+
from keras.optimizers import Adagrad
|
|
36
|
+
|
|
37
|
+
import matplotlib.pyplot as plt
|
|
38
|
+
|
|
39
|
+
%matplotlib inline
|
|
40
|
+
|
|
41
|
+
|
|
42
|
+
|
|
43
|
+
data = pd.read_csv('international-airline-passengers.csv', skipfooter=3)
|
|
44
|
+
|
|
45
|
+
data = data.values[:, -1]
|
|
46
|
+
|
|
47
|
+
data = data[1:] - data[:-1]
|
|
48
|
+
|
|
49
|
+
|
|
50
|
+
|
|
51
|
+
inputdata = []
|
|
52
|
+
|
|
53
|
+
target = []
|
|
54
|
+
|
|
55
|
+
input_len = 5
|
|
56
|
+
|
|
57
|
+
|
|
58
|
+
|
|
59
|
+
for i in range(0, len(data)-input_len):
|
|
60
|
+
|
|
61
|
+
inputdata.append(data[i:i+input_len])
|
|
62
|
+
|
|
63
|
+
target.append(data[i+input_len])
|
|
64
|
+
|
|
65
|
+
|
|
66
|
+
|
|
67
|
+
from sklearn.model_selection import train_test_split
|
|
68
|
+
|
|
69
|
+
|
|
70
|
+
|
|
71
|
+
X = np.array(inputdata).reshape(-1, input_len, 1)
|
|
72
|
+
|
|
73
|
+
y = np.array(target).reshape(-1, 1)
|
|
74
|
+
|
|
75
|
+
(train_X, test_X, train_y, test_y) = train_test_split(X, y, test_size=0.2, shuffle=False)
|
|
76
|
+
|
|
77
|
+
|
|
78
|
+
|
|
79
|
+
n_in = 1
|
|
80
|
+
|
|
81
|
+
n_hidden = 50
|
|
82
|
+
|
|
83
|
+
n_out = 1
|
|
84
|
+
|
|
85
|
+
model = Sequential()
|
|
86
|
+
|
|
87
|
+
model.add(LSTM(n_hidden, input_shape=(input_len,n_in)))
|
|
88
|
+
|
|
89
|
+
model.add(Dense(100, activation='relu'))
|
|
90
|
+
|
|
91
|
+
model.add(Dropout(0.5))
|
|
92
|
+
|
|
93
|
+
model.add(Dense(5, activation='relu'))
|
|
94
|
+
|
|
95
|
+
model.add(Dense(100, activation='relu'))
|
|
96
|
+
|
|
97
|
+
model.add(Dropout(0.5))
|
|
98
|
+
|
|
99
|
+
model.add(Dense(n_out, activation='linear'))
|
|
100
|
+
|
|
101
|
+
opt = Adagrad()
|
|
102
|
+
|
|
103
|
+
model.compile(loss="mean_squared_error",optimizer=opt)
|
|
104
|
+
|
|
105
|
+
|
|
106
|
+
|
|
107
|
+
model.fit(train_X, train_y, batch_size=len(X)//10, epochs=50, validation_data=(test_X, test_y))
|
|
108
|
+
|
|
109
|
+
|
|
110
|
+
|
|
111
|
+
py = model.predict(train_X)
|
|
112
|
+
|
|
113
|
+
py_ = model.predict(test_X)
|
|
114
|
+
|
|
115
|
+
|
|
116
|
+
|
|
117
|
+
px = np.arange(predicted.shape[0])
|
|
118
|
+
|
|
119
|
+
|
|
120
|
+
|
|
121
|
+
fig, ax = plt.subplots(dpi=200)
|
|
122
|
+
|
|
123
|
+
ax.plot(y, label="original")
|
|
124
|
+
|
|
125
|
+
ax.plot(px[:len(py)], py, label="predicted_train", color='orange')
|
|
126
|
+
|
|
127
|
+
ax.plot(px[len(py):], py_, label="predicted_test", color='red')
|
|
128
|
+
|
|
129
|
+
plt.legend()
|
|
130
|
+
|
|
131
|
+
plt.grid()
|
|
132
|
+
|
|
133
|
+
plt.show()
|
|
134
|
+
|
|
135
|
+
```
|