-
Notifications
You must be signed in to change notification settings - Fork 0
/
time predict_LSTM.py
176 lines (143 loc) · 5.41 KB
/
time predict_LSTM.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
import numpy as np
import pandas as pd
from predict_data import PredictData
from fill_normalization import fill, normalization
import matplotlib.pyplot as plt
from keras.models import Sequential
from keras.layers import LSTM, Bidirectional, Dense, Dropout
from keras.callbacks import EarlyStopping
from keras.optimizers import Adam
from keras.models import load_model
from sklearn.preprocessing import MinMaxScaler
from keras.utils.vis_utils import plot_model
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence) - 1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return np.array(X), np.array(y)
def exponential_smoothing(alpha, s):
'''
一次指数平滑
:param alpha: 平滑系数
:param s: 数据序列， list
:return: 返回一次指数平滑模型参数， list
'''
s_temp = []
s_temp.append(s[0])
for i in range(1, len(s), 1):
s_temp.append(alpha * s[i-1] + (1 - alpha) * s_temp[i-1])
return s_temp
data = PredictData()
data = fill(data, 84, 336, 12)
data = data.reshape(2016, 14)
c = data[:, 1]
np.random.seed(5)
a = np.random.randint(0, 12)
c = c[168*a: 168*(a+1)]
# print(c)
# print(type(c)) #numpy.ndarray
# print(c.shape) # (168,)
time = pd.Series(np.array(c, dtype=float), index=pd.date_range(start='2018-10-08', periods=168, freq='H'))
# print(time)
dict_time = {'ds': time.index, 'y': time.values}
df_time = pd.DataFrame(dict_time)
# 去掉异常值
td = df_time['y'].describe() # 描述性统计得到：min，25%，50%，75%，max值
high = td['75%'] + 1.5 * (td['75%'] - td['25%']) # 定义高点阈值，1.5倍四分位距之外
low = td['25%'] - 1.5 * (td['75%'] - td['25%']) # 定义低点阈值
forbid_index = df_time['y'][(df_time['y'] > high) | (df_time['y'] < low)].index # 变化幅度超过阈值的点的索引
li = list(df_time['y'])
for i in range(len(forbid_index)):
li[forbid_index[i]] = (li[forbid_index[i]-1]+li[forbid_index[i]+1])/2
df_time['y'] = np.array(li)
# df_time['y'].plot()
df_time['y'] = exponential_smoothing(0.6, list(df_time['y']))
# df_time['y'].plot()
# plt.show()
raw_seq = df_time['y'].to_list()
# n_steps = 10
n_steps = 20
n_features = 1
X, y = split_sequence(raw_seq, n_steps)
# print(X)
# print(y)
"""
LSTM对输入数据的规模很敏感，特别是当使用sigmoid（默认）或tanh激活功能时。
将数据重新调整到0到1的范围（也称为标准化）可能是一个很好的做法。
我们可以使用scikit-learn库中的MinMaxScaler预处理类轻松地规范数据集。
使用MinMaxScaler().inverse_transform可以恢复数据
"""
scalex = MinMaxScaler(feature_range=(0, 1))
scaley = MinMaxScaler(feature_range=(0, 1))
X = scalex.fit_transform(X)
y = scaley.fit_transform(y.reshape(y.shape[0], 1))
X = X.reshape((X.shape[0], X.shape[1], n_features))
split = int(0.8 * len(X))
x_train = X[:split]
x_test = X[split:]
y_train = y[:split].flatten()
y_test = y[split:].flatten()
# # define model
# model = Sequential()
# model.add(LSTM(100, activation='relu', return_sequences=True, input_shape=(n_steps, n_features)))
# model.add(Dropout(0.1))
# model.add(LSTM(100, activation='relu', return_sequences=True))
# model.add(Dropout(0.1))
# model.add(LSTM(100, activation='relu'))
# # model.add(Dropout(0.2))
# model.add(Dense(1))
# model.compile(optimizer='adam', loss='mse')
# # early_stop = EarlyStopping(monitor='loss', patience=10)
# # fit model
# history = model.fit(x_train, y_train, epochs=400, verbose=1, batch_size=10)
model = Sequential()
model.add(LSTM(100, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# early_stop = EarlyStopping(monitor='loss', patience=10)
# fit model
history = model.fit(x_train, y_train, epochs=800, verbose=1)
model.save("LSTM-.h5")
# plt.plot(history.history['loss'], label='loss')
# plt.title('LOSS')
# plt.ylabel('loss')
# plt.xlabel('epoch')
# plt.legend()
# plt.show()
#
# plot_model(model, to_file='model-LSTM.png', show_shapes=True)
model = load_model("LSTM-.h5")
score_lstm = model.evaluate(x_train, y_train)
print('LSTM:', score_lstm)
plt.subplot2grid((1, 2), (0, 0))
pred_train = model.predict(x_train)
inv_y_train = scaley.inverse_transform(y_train.reshape(y_train.shape[0], 1)).flatten()
inv_pred = scaley.inverse_transform(pred_train.reshape(y_train.shape[0], 1)).flatten()
plt.plot(pred_train, label='pred_train')
plt.plot(y_train, label='true_train')
# plt.plot(inv_pred, label='inv_pred')
# plt.plot(inv_y_train, label='inv_y_train')
plt.legend()
# plt.show()
score_lstm = model.evaluate(x_test, y_test)
print('LSTM:', score_lstm)
plt.subplot2grid((1, 2), (0, 1))
pred_test = model.predict(x_test)
plt.plot(pred_test, label='pred_test')
plt.plot(y_test, label='true_test')
plt.legend()
plt.show()
"""
问题1：过拟合严重，暂时解决不了，可能还是需要继续调参，包括时间步和dropout和LSTM层数
问题2：这只是对单个特征做预测,已解决
问题3：最小最大归一化的恢复问题，已解决
"""