-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtype_2.py
206 lines (179 loc) · 7.07 KB
/
type_2.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
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import GoogleTrend
import RoadSection
import bpnn
import dynamic_bpnn
import time
from sklearn.metrics import mean_squared_error
def normalize(series, x):
return 2 / (series.max() - series.min()) * (x - series.min()) - 1
def normalize_all(series):
series_max = series.max()
series_min = series.min()
return pd.Series([(lambda x: 2 / (series_max - series_min) * (x - series_min) - 1)(x) for x in series])
def anti_normalize_all(src_series, target_series):
src_series_max = src_series.max()
src_series_min = src_series.min()
return pd.Series(
[(lambda y: (y + 1) * ((src_series_max - src_series_min) / 2) + src_series_min)(y) for y in target_series])
# Load Data-----------------------------------------------------
rs = RoadSection.RoadSection('03F2614S-03F2709S.csv')
rs.load()
# rs.normalize()
gt = GoogleTrend.GoogleTrend('multiTimeline.csv')
gt.load()
# gt.normalize()
print(len(rs.flow_all))
print(len(gt.trend_percentage))
assert rs.really_start_day_index == gt.trend_start_day_index # 416
# Store to Pandas DataFrame Type-------------------------------------------
# dates = pd.date_range('20150101', periods=639)
df = pd.DataFrame({
'flow': rs.flow_all, # df[:, 0]
'trend': gt.trend_percentage # df[:, 1]
},
# index=dates
)
flow_week_dict = {
'0': rs.flow_all[rs.really_start_day_index::7],
'1': rs.flow_all[rs.really_start_day_index + 1::7],
'2': rs.flow_all[rs.really_start_day_index + 2::7],
'3': rs.flow_all[rs.really_start_day_index + 3::7],
'4': rs.flow_all[rs.really_start_day_index + 4::7],
'5': rs.flow_all[rs.really_start_day_index + 5::7],
'6': rs.flow_all[rs.really_start_day_index + 6::7]
}
trend_week_dict = {
'0': gt.trend_percentage[rs.really_start_day_index::7],
'1': gt.trend_percentage[rs.really_start_day_index + 1::7],
'2': gt.trend_percentage[rs.really_start_day_index + 2::7],
'3': gt.trend_percentage[rs.really_start_day_index + 3::7],
'4': gt.trend_percentage[rs.really_start_day_index + 4::7],
'5': gt.trend_percentage[rs.really_start_day_index + 5::7],
'6': gt.trend_percentage[rs.really_start_day_index + 6::7]
}
df_flow_week = pd.DataFrame.from_dict(flow_week_dict, orient='index').T
df_trend_week = pd.DataFrame.from_dict(trend_week_dict, orient='index').T
print(df)
print(df_flow_week)
print(df_trend_week)
# exit()
# Type 2 Neural Network---------------------------------------------------------------
df_flow_week_sub_mean = df_flow_week.copy()
df_trend_week_sub_mean = df_trend_week.copy()
assert df_flow_week_sub_mean.shape == df_trend_week_sub_mean.shape
# subtract mean
for col in range(df_flow_week_sub_mean.shape[1]):
df_flow_week_sub_mean.iloc[:, col] -= df_flow_week.iloc[:, col].mean()
df_trend_week_sub_mean.iloc[:, col] -= df_trend_week.iloc[:, col].mean()
print(df_flow_week_sub_mean)
print(df_trend_week_sub_mean)
# transform to series
series_flow_week_sub_mean = pd.Series(pd.concat(
[(lambda row: df_flow_week_sub_mean.iloc[row, :])(row) for row in range(df_flow_week_sub_mean.shape[0])], axis=0,
ignore_index=True))
print(series_flow_week_sub_mean)
series_trend_week_sub_mean = pd.concat(
[(lambda row: df_trend_week_sub_mean.iloc[row, :])(row) for row in range(df_trend_week_sub_mean.shape[0])], axis=0,
ignore_index=True)
print(series_trend_week_sub_mean)
# normalization
point = 10 # 先跳過有問題那幾筆
series_flow_week_sub_mean_normalization = normalize_all(series_flow_week_sub_mean[point:])
series_trend_week_sub_mean_normalization = normalize_all(series_trend_week_sub_mean[point:])
print('***', len(series_flow_week_sub_mean_normalization), point)
df_FnT_week_sub_mean_normalization = pd.DataFrame({
'f_sn': series_flow_week_sub_mean_normalization,
't_sn': series_trend_week_sub_mean_normalization
})
print(df_FnT_week_sub_mean_normalization)
# plt.plot(df_FnT_week_sub_mean_normalization)
# plt.show()
# exit()
# anti-normalization
series_flow_week_sub_mean_anti_normalization = anti_normalize_all(series_flow_week_sub_mean[point:],
series_flow_week_sub_mean_normalization)
series_trend_week_sub_mean_anti_normalization = anti_normalize_all(series_trend_week_sub_mean[point:],
series_trend_week_sub_mean_normalization)
print('anti!!!!!!!!!!!!!!!!!!!!!!!')
cp_series_flow_week_sub_mean = series_flow_week_sub_mean[point:].copy()
cp_series_trend_week_sub_mean = series_trend_week_sub_mean[point:].copy()
print(type(series_flow_week_sub_mean), type(series_trend_week_sub_mean))
df_FnT_week_sub_mean_normalization = pd.DataFrame({
'f': cp_series_flow_week_sub_mean,
't': cp_series_trend_week_sub_mean,
'f_san': series_flow_week_sub_mean_anti_normalization,
't_san': series_trend_week_sub_mean_anti_normalization
})
print(df_FnT_week_sub_mean_normalization)
# exit()
cases = []
labels = []
for i in range(0, 100):
five_days = list(series_flow_week_sub_mean_normalization[i:i + 5]) + list(
series_trend_week_sub_mean_normalization[i:i + 5])
cases.append(five_days)
print(i, i + 5, [series_flow_week_sub_mean_normalization[i + 5]])
labels.append([series_flow_week_sub_mean_normalization[i + 5]])
cases_test = []
labels_test = []
google = []
for i in range(100, 200):
five_days = list(series_flow_week_sub_mean_normalization[i:i + 5]) + list(
series_trend_week_sub_mean_normalization[i:i + 5])
cases_test.append(five_days)
print(i, i + 5, [series_flow_week_sub_mean_normalization[i + 5]])
labels_test.append([series_flow_week_sub_mean_normalization[i + 5]])
google.append([series_trend_week_sub_mean_normalization[i + 5]])
# exit()
# # Basic version
# nn = bpnn.BPNeuralNetwork()
# nn.setup(10, 2, 1)
# while True:
# start = time.time()
# nn.train(cases_test, labels_test)
# predict_all = nn.test(cases, labels)
# end = time.time()
# elapsed = end - start
# print("Time taken: ", elapsed, "seconds.")
# if nn.mse < 0.21:
# break
# # TensorFlow version
# nn = dynamic_bpnn.BPNeuralNetwork()
# nn.setup(10, [2], 1)
# while True:
# start = time.time()
# nn.train(cases, labels)
# predict_all = nn.test(cases_test, labels_test)
# end = time.time()
# elapsed = end - start
# print("Time taken: ", elapsed, "seconds.")
# if nn.mse < 0.020:
# break
# nn.save_model('tf_model\save_net.ckpt')
# Load
nn = dynamic_bpnn.BPNeuralNetwork()
nn.setup(10, [2], 1)
nn.load_model('tf_model\save_net.ckpt')
predict_all = nn.test(cases_test, labels_test)
plt.plot(labels_test, 'b')
plt.plot(predict_all, 'r')
# plt.plot(google)
plt.show()
# predict_all = nn.test(cases, labels)
# plt.plot(labels, 'b')
# plt.plot(predict_all, 'r')
# # plt.plot(google)
# plt.show()
# print(type(labels_test))
# print(labels_test)
# # exit()
# b = anti_normalize_all(series_flow_week_sub_mean, [item[0] for item in labels_test])
# p = anti_normalize_all(series_flow_week_sub_mean, predict_all)
# plt.plot(b, 'b')
# plt.plot(p, 'r')
# # plt.plot(google)
# plt.show()
# print(mean_squared_error(b, p))