-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmrn_test.py
295 lines (274 loc) · 11.1 KB
/
mrn_test.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
#两个prior模型的输出,for循环
import torch.optim.lr_scheduler as lr_scheduler
from torch.utils.data import Dataset, DataLoader, TensorDataset
import torch.nn as nn
import torchvision.transforms as transforms
from sklearn.preprocessing import MinMaxScaler
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import os
import random
os.environ['PYTHONHASHSEED'] = '0'
seed = 14
np.random.seed(seed)
random.seed(seed)
import torch
from torch.autograd import Variable
import tensor_op
import time
torch.manual_seed(6)
# Hyper Parameters
INPUT_SIZE = 7 # input size（输入特征维度）
TIME_STEPS = 6 # time step（步长：使用前TIME_STEP 条数据进行预测）
batch_size = 288
cov_update_freq=50
class_cov_num=4
# t = 0
# s = 0
#验证时，t设为0，s设为1到s_max；测试时，s设为0，t设为0-29
model_path="model_result/mrnModel3_epo180.pth"
# divide ts as training/testing samples,looBack is lag window(time step)
def createSamples(dataset, lookBack, RNN=True):
dataX, dataY = [], []
for i in range(len(dataset) - lookBack):
sample_X = dataset[i:(i + lookBack), :]
sample_Y = dataset[i + lookBack, :]
dataX.append(sample_X)
dataY.append(sample_Y)
dataX = np.array(dataX) # (N, lag, 1)
dataY = np.array(dataY) # (N, 1)
if not RNN:
dataX = np.reshape(dataX, (dataX.shape[0], dataX.shape[1]))
return dataX, dataY
# 加载数据
data = pd.read_csv("20151101cba-mtl2.csv")
data = data.drop(['Datetime'], axis=1)
data = np.array(data, dtype='float32')
flow = data[:, 0] # drop之后的第0列
speed = data[:, 1]
scaler1 = MinMaxScaler()
scaler2 = MinMaxScaler()
scaler3 = MinMaxScaler()
data = scaler1.fit_transform(data)
flow = flow.reshape(-1, 1)
speed = speed.reshape(-1, 1)
flow = scaler2.fit_transform(flow)
speed = scaler3.fit_transform(speed)
flow_data = data[:, 0].reshape(len(data), 1)
speed_data = data[:, 1].reshape(len(data), 1)
##create dataset
data_X, _ = createSamples(data, TIME_STEPS)
_, data_f = createSamples(flow_data, TIME_STEPS)
_, data_s = createSamples(speed_data, TIME_STEPS)
length = 396 * 288 -288*153-TIME_STEPS #153对应到6月
class Time_Series_Data(Dataset):
def __init__(self, train_x, train_y1, train_y2):
self.X = train_x
self.y1 = train_y1
self.y2 = train_y2
def __getitem__(self, item):
x_t = self.X[item]
y_t1 = self.y1[item]
y_t2 = self.y2[item]
return x_t, y_t1, y_t2
def __len__(self):
return len(self.X)
# testdataset = Time_Series_Data(test_X, test_f, test_s)
# testloader = DataLoader(testdataset, batch_size=batch_size, shuffle=False)
class MRNmodel(nn.Module):
def __init__(self):
super(MRNmodel, self).__init__()
self.conv1 = nn.Sequential(
nn.Conv1d(in_channels=INPUT_SIZE, out_channels=64, kernel_size=1),
nn.ReLU(),
nn.Dropout(0.5),
)
self.linear=nn.Linear(64,64)
self.bilstmf = nn.LSTM(
# input_size=units*TIME_STEPS,
input_size=64,
hidden_size=32,
num_layers=1,
batch_first=True,
bidirectional=True # bilstm
)
self.bilstms = nn.LSTM(
# input_size=64*TIME_STEPS,
input_size=64,
hidden_size=32,
num_layers=1,
batch_first=True,
bidirectional=True # bilstm
)
self.dropoutf = nn.Dropout(0.5)
self.dropouts = nn.Dropout(0.5)
# self.convf = nn.Sequential(
# nn.Conv1d(in_channels=128, out_channels=64, kernel_size=3),
# nn.ReLU(),
# nn.Dropout(0.5),
# )
# self.convs = nn.Sequential(
# nn.Conv1d(in_channels=128, out_channels=64, kernel_size=3),
# nn.ReLU(),
# nn.Dropout(0.5),
# )
# self.linear=nn.Linear(128,128)
# self.linearf = nn.Linear(2*64,128) #bilstm应该2*hidden_size
# self.linears = nn.Linear(2*64,128)
self.linearf1 = nn.Linear(32*2,class_cov_num) # bilstm应该2*hidden_size
self.linears1 = nn.Linear(32*2,class_cov_num)
self.reluf1 = nn.ReLU()
self.relus1 = nn.ReLU()
self.linearf2 = nn.Linear(class_cov_num*TIME_STEPS,1)
self.linears2 = nn.Linear(class_cov_num*TIME_STEPS,1)
nn.init.xavier_uniform_(self.linearf1.weight, gain=1)
nn.init.xavier_uniform_(self.linears1.weight, gain=1)
self.linearf1.bias.data.fill_(0.0)
self.linears1.bias.data.fill_(0.0)
nn.init.xavier_uniform_(self.linearf2.weight, gain=1)
nn.init.xavier_uniform_(self.linears2.weight, gain=1)
self.linearf2.bias.data.fill_(0.0)
self.linears2.bias.data.fill_(0.0)
self.hidden_cellf = None
self.hidden_cells = None
self.trade_off = 1
self.bottleneck_size = self.linearf1.in_features
self.bottleneck_size1 = self.linearf2.in_features
self.iter_num = 0
self.cov_update_freq = cov_update_freq
if self.trade_off > 0:
# initialize covariance matrix
self.task_cov = torch.eye(2) # E(2)
self.class_cov = torch.eye(class_cov_num) # E(8)
self.feature_cov = torch.eye(self.bottleneck_size) # 生成128*128的单位矩阵E(128)
self.task_cov_var = Variable(self.task_cov)
self.class_cov_var = Variable(self.class_cov)
self.feature_cov_var = Variable(self.feature_cov)
self.task_cov1 = torch.eye(2) # E(2)
self.class_cov1 = torch.eye(1) # E(1)
self.feature_cov1 = torch.eye(self.bottleneck_size1) # 生成128*128的单位矩阵E(128)
self.task_cov_var1 = Variable(self.task_cov1)
self.class_cov_var1 = Variable(self.class_cov1)
self.feature_cov_var1 = Variable(self.feature_cov1)
def forward(self, x):
# print(x.size())
convout = self.conv1(x.transpose(1, 2))
# print(convout.size())
convout = self.linear(convout.view(len(convout),TIME_STEPS,-1))
# print(convout.size())
lstm_outf, self.hidden_cellf = self.bilstmf(convout.view(len(convout),TIME_STEPS,-1),self.hidden_cellf)
lstm_outs, self.hidden_cells = self.bilstms(convout.view(len(convout),TIME_STEPS,-1),self.hidden_cells)
lstm_outf = self.dropoutf(lstm_outf)
lstm_outs = self.dropouts(lstm_outs)
# lstm_outf = self.convf(lstm_outf.transpose(1, 2))
# lstm_outs = self.convs(lstm_outs.transpose(1, 2))
fc_outf = self.linearf1(lstm_outf.view(len(lstm_outf),TIME_STEPS,-1))
fc_outs = self.linears1(lstm_outs.view(len(lstm_outf),TIME_STEPS,-1))
fc_outf = self.reluf1(fc_outf)
fc_outs = self.relus1(fc_outs)
predf = self.linearf2(fc_outf.view(len(convout), -1)) # .view(len(convout), -1)
preds = self.linears2(fc_outs.view(len(convout), -1))
return predf, preds
def mape(y_true, y_pred):
n = len(y_true)
mape = sum(np.abs((y_true - y_pred) / y_true)) / n * 100
return mape
def se(y_true,y_pred):
n=len(y_true)
se=sum(((y_true-y_pred)/y_true-mape(y_true,y_pred))**2)/n
return se
model = MRNmodel()
model.load_state_dict(torch.load(model_path))
mape_f,mape_s=[],[]
#predict
model.hidden_cellf = None
model.hidden_cells = None
mapef=0
mapes=0
for t in range(0,30):
test_X = data_X[length - 288 * (30 - t):length - 288 * (29 - t), :, :]
test_f = data_f[length - 288 * (30 - t):length - 288 * (29 - t)]
test_s = data_s[length - 288 * (30 - t):length - 288 * (29 - t)]
testdataset = Time_Series_Data(test_X, test_f, test_s)
testloader = DataLoader(testdataset, batch_size=batch_size, shuffle=False)
##predict,预测
model.eval()
# model.hidden_cellf=None
# model.hidden_cells=None
pred_f, pred_s = [], []
for j, (X_, yf_,ys_) in enumerate(testloader):
model.hidden_cellf = None
model.hidden_cells = None
X_ = Variable(X_)
# print('---------------------------')
# print(X_)
predf,preds = model(X_)
predf,preds = predf.data.numpy(),preds.data.numpy()
predf,preds = np.squeeze(predf),np.squeeze(preds)
pred_f.append(predf)
pred_s.append(preds)
pred_f,pred_s=np.array(pred_f),np.array(pred_s)
pred_f,pred_s=pred_f.reshape(288,1),pred_s.reshape(288,1)
pred_f = scaler2.inverse_transform(pred_f)
test_f = scaler2.inverse_transform(test_f)
pred_s = scaler3.inverse_transform(pred_s)
test_s = scaler3.inverse_transform(test_s)
mapef+=mape(test_f[72:264], pred_f[72:264])
mapes+=mape(test_s[72:264], pred_s[72:264])
mape_f.append(mape(test_f[72:264], pred_f[72:264]))
mape_s.append(mape(test_s[72:264], pred_s[72:264]))
print('flow MAPE:', mape(test_f[72:264], pred_f[72:264])) # test_f[72:264], pred_f[72:264]
print('speed MAPE:', mape(test_s[72:264], pred_s[72:264])) # test_s[72:264], pred_s[72:264]
# if t==7:
# fig1 = plt.figure()
# ax1 = fig1.add_subplot(111)
# plt.title('traffic flow prediction')
# plt.plot(test_f, 'b', label='real flow')
# plt.plot(pred_f, 'r', label='predict flow')
# plt.legend(loc='upper left')
# xticks = [0, 48, 96, 144, 192, 240, 288]
# xlabels = ['0:00', '4:00', '8:00', '12:00', '16:00', '20:00', '24:00']
# ax1.set_xticks(xticks)
# ax1.set_xticklabels(xlabels) # , rotation='45') #rotation='vertical'
# plt.xlim(0, 288)
# plt.show()
print("Average flow mape:",mapef/30," Average speed mape:",mapes/30)
# print("Average speed mape:",mapes/30)
mape_f=pd.DataFrame(mape_f)
mape_s=pd.DataFrame(mape_s)
mape_f.to_csv("model_result/mape_f.csv")
mape_s.to_csv("model_result/mape_s.csv")
# print('pred_f[:10]', pred_f[:10], 'pred_f[-10:]', pred_f[-10:])
# print('test_f[:10]', test_f[:10], 'test_f[-10:]', test_f[-10:])
# print('pred_s[:10]', pred_s[:10], 'pred_s[-10:]', pred_s[-10:])
# fig1 = plt.figure()
# ax1 = fig1.add_subplot(111)
# plt.title('traffic flow prediction')
# plt.plot(test_f, 'b', label='real flow')
# plt.plot(pred_f, 'r', label='predict flow')
# plt.legend(loc='upper left')
# xticks = [0, 48, 96, 144, 192, 240, 288]
# xlabels = ['0:00', '4:00', '8:00', '12:00', '16:00', '20:00', '24:00']
# ax1.set_xticks(xticks)
# ax1.set_xticklabels(xlabels) # , rotation='45') #rotation='vertical'
# plt.xlim(0, 288)
# plt.ylim(0, 530)
#
# # plt.figure()
# # plt.title('difference flow prediction')
# # plt.plot(test_f[0:287], 'b', label='real flow')
# # plt.plot(pred_f[1:288], 'r', label='predict flow')
# # plt.legend(loc='upper left')
#
# fig2 = plt.figure()
# ax2 = fig2.add_subplot(111)
# plt.title('traffic speed prediction')
# plt.plot(test_s, 'b', label='real speed')
# plt.plot(pred_s, 'r', label='predict speed')
# plt.legend(loc='upper left')
# ax2.set_xticks(xticks)
# ax2.set_xticklabels(xlabels)
# plt.xlim(0, 288)
# plt.ylim(min(test_s) - 20, max(test_s + 20))
# plt.show()