-
Notifications
You must be signed in to change notification settings - Fork 0
/
LSTM2_simple_14_04.py
470 lines (346 loc) · 14.1 KB
/
LSTM2_simple_14_04.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
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
#!/usr/bin/env python
# coding: utf-8
# sys.path.insert(0,'C:\\Users\gneven\AppData\Roaming\Python\Python38\site-packages')
# In[1]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import datetime
import time
import multiprocessing as mp
from itertools import product
from keras.models import Sequential,Model
from keras.layers import *#Dense, Input, Concatenate, Lambda,LSTM,Dropout,Layer,Flatten
from keras.layers import ConvLSTM2D,BatchNormalization,TimeDistributed,MaxPooling3D
from keras.optimizers import RMSprop, Adam
from keras import optimizers
from keras.utils.vis_utils import plot_model
import keras.losses as kl
import pydot
from keras import optimizers
import keras
import pickle
''
import tensorflow as tf
import sys
import warnings
import os
#from tqdm import tqdm
os.chdir('/gneven/Desktop/ETH/Sem2/Projeckt/Pyhton/')
from scipy.interpolate import UnivariateSpline
from functions import *
idx = pd.IndexSlice
tf.config.threading.set_intra_op_parallelism_threads(4)
tf.config.threading.set_inter_op_parallelism_threads(4)
# In[3]:
def isgood(day):
Data_d = Data.loc[(Data['linie'] == Studied_line) & (Data['datum_nach'] == day),:]
Data_d['Tot_latness'] = Data_d['ist_ab_nach']-Data_d['soll_ab_nach']
idBus = Data_d.loc[(Data_d['soll_ab_nach']>=startTime ) &(
(Data_d['soll_ab_nach']<=endTime )
) & (Data_d['halt_diva_nach'] == 1143) ,'fahrzeug'].unique()
if idBus.shape[0] != 13:
return False
else:
return True
def smoothing(y):
y = np.asarray(y,dtype=np.float64)
x = np.asarray([range(y.shape[0])])
w = np.isnan(y)
y[w] = 0.
spl = UnivariateSpline(x, y, w=~w)
spl.set_smoothing_factor(50)
return spl(x)
before_dt = 20
dt = 60
startTime = 7*3600
endTime = 10*3600 + before_dt*dt
ncoeur = 4
allfile = os.listdir('data/data2/Times/')
Data = get_time_data('data/data2/Times/'+allfile[0])
#for str in allfile[1:]:
# Data = pd.concat([Data,get_time_data('data/data2/Times/'+str)])
#
Data = Data.loc[Data['datum_nach'].dt.dayofweek < 5,:]
#delete depot run
Data = Data[~Data['fw_lang'].str.contains('DEP|Einfahrt|Ausfahrt')]
Studied_line = 32
Data = Data[Data['linie']<1000]
Data = Data.sort_values(by ='halt_diva_nach')
Data['ID'] = 100000000*Data['richtung']+ 10000*Data['linie']+Data['halt_diva_nach']
Data = Data.sort_values(by ='ID')
i = 0
for day in Data['datum_nach'].unique():
if isgood(day):
if i == 0:
splitData = list([Data.loc[Data['datum_nach'] == day,:]])
else:
splitData.append(Data.loc[Data['datum_nach'] == day,:])
i+=1
# ## Selection
# ### Stops
# Select the 3 lateness (travel time, stops and total ) for every stops every 30s, if no new bus came, the last one is kept. It was first thought to take as time step the updating of the bus lateness, but as we now have multiple bus it does not work.
# In[ ]:
t1 = time.time()
#Data_d = splitData[0]
def features_selector(Data_d):
#for i in [1]:
day = Data_d['datum_nach'].unique().astype('int')
Data_d = pd.DataFrame({
'Delta_stops' :(Data_d['ist_ab_nach']-Data_d['ist_an_nach1']-Data_d['soll_ab_nach']+Data_d['soll_an_nach']),
'Delta_trip' :(Data_d['ist_an_nach1']-Data_d['ist_ab_von']-Data_d['soll_an_nach']+Data_d['soll_ab_von']),
'Tot_lat' :Data_d['ist_ab_nach']-Data_d['soll_ab_nach'],
'ID' :Data_d['ID'],
'time' :Data_d['ist_ab_nach']})
#
# colname = [np.repeat(Data['ID'].unique(),3),np.resize(['Delta_stops','Delta_trip','Tot_lat'
# ],3*Data['ID'].unique().shape[0])]
out_S = pd.DataFrame(index = range(startTime,endTime,dt) ,columns= Data['ID'].unique()) #+before_dt*dt
out_T = pd.DataFrame(index = range(startTime,endTime,dt) ,columns= Data['ID'].unique()) #+before_dt*dt
out_TOT = pd.DataFrame(index = range(startTime,endTime,dt) ,columns= Data['ID'].unique())#+before_dt*dt
t_last = startTime
Data_t = Data_d.loc[(Data_d['time'] < t_last) ,['Delta_stops','Delta_trip','Tot_lat','ID']]
# out.loc[t_last,idx[Data_t['ID'].values,:]] = Data_t.groupby('ID').last().values.ravel()
# out.loc[t,:].update(Data_t.groupby('fahrzeug').last()['Tot_latness'])
out_S.loc[t_last,:].update( Data_t.groupby('ID').last()['Delta_stops'])
out_T.loc[t_last,:].update( Data_t.groupby('ID').last()['Delta_trip'])
out_TOT.loc[t_last,:].update( Data_t.groupby('ID').last()['Tot_lat'])
for t in range(startTime+dt,endTime,dt):
Data_t = Data_d.loc[(Data_d['time'] > t_last) &(Data_d['time'] <= t),['Delta_stops','Delta_trip','Tot_lat','ID']]
out_S.loc[t,:] = out_S.loc[t_last,:]
out_T.loc[t,:] = out_T.loc[t_last,:]
out_TOT.loc[t,:] = out_TOT.loc[t_last,:]
out_S.loc[t,:].update( Data_t.groupby('ID').last()['Delta_stops'])
out_T.loc[t,:].update( Data_t.groupby('ID').last()['Delta_trip'])
out_TOT.loc[t,:].update( Data_t.groupby('ID').last()['Tot_lat'])
t_last = t
# out_S = out_S.dropna(axis='columns',how='all')
# out_T = out_T.dropna(axis='columns',how='all')
# out_TOT = out_TOT.dropna(axis='columns',how='all')
# out_S = out_S.apply(smoothing, axis=0)
# out_T = out_T.apply(smoothing, axis=0)
# out_TOT = out_TOT.apply(smoothing, axis=0)
out = np.array([out_S.to_numpy(),out_T.to_numpy(),out_TOT.to_numpy()])
return_array = np.empty(((int((endTime-startTime)/dt)),out.shape[0],before_dt,out.shape[2]))
for w in range(int((endTime-startTime)/dt)-before_dt):
return_array[w,:,:,:] = out[:,w:(before_dt+w),:]
np.save((os.getcwd()+'/DataPickles/FrameT%s' %day),return_array)
return return_array
del(Data_t,Data_d,out,return_array,out_S,out_T,out_TOT)
pool = mp.Pool(ncoeur)
#i = 0
#for day in Data['datum_nach'].unique():
# if i == 0:
# splitData = list([Data.loc[Data['datum_nach'] == day,:]])
# else:
# splitData.append(Data.loc[Data['datum_nach'] == day,:])
# i+=1
FrameT = np.array(pool.map(features_selector,splitData),np.float32)
pool.close()
pool.join() # 28
print('Time spent: ',time.time()-t1)
# ### Reshape into (#sample , time ,features, #stop, 1)
# In[ ]:
FrameT = np.moveaxis(FrameT, [0, 1, 2, 3,4], [0, 1, 3, 2,4])
FrameT = np.expand_dims(FrameT,axis = 5)
FrameT.shape #(5, 360, 3, 2472, 1) 1539
#FrameT = FrameT[:,:,:,:,~np.all(np.isnan(FrameT),axis = (0,1,2,3,5)),:]
FrameT[:,:,:,0,:,:] = FrameT[:,:,:,0,:,:]/np.nanmax(abs(FrameT[:,:,:,0,:,:]))
FrameT[:,:,:,1,:,:] = FrameT[:,:,:,1,:,:]/np.nanmax(abs(FrameT[:,:,:,1,:,:]))
FrameT[:,:,:,2,:,:] = FrameT[:,:,:,2,:,:]/np.nanmax(abs(FrameT[:,:,:,2,:,:]))
# ### Bus
# The bus delay for all the bus in the line is computed every 30s. ( don't need reshape already good)
# In[ ]:
num_next_time = 10
nex_time_dt = 1
startTime += before_dt*dt
endTime += before_dt*dt
#Data_d = splitData[0]
#def get_out(Data_d):
out_fin = []
for Data_d in splitData:
Data_d = Data_d.loc[Data_d['linie'] == Studied_line,:]
Data_d['Tot_latness'] = Data_d['ist_ab_nach']-Data_d['soll_ab_nach']
idBus = Data_d.loc[(Data_d['soll_ab_nach']>=startTime ) &(
(Data_d['soll_ab_nach']<=endTime )
) & (Data_d['halt_diva_nach'] == 1143) ,'fahrzeug'].unique()
if idBus.shape[0] != 13:
print('wtf',Data_d['datum_nach'].dt.dayofweek.unique(),Data_d['datum_nach'].unique())
else:
out= pd.DataFrame(index = range(startTime,endTime+num_next_time*nex_time_dt*dt,dt), columns = idBus)
t_last = startTime
Data_t = Data_d.loc[(Data_d['ist_ab_nach'] < t_last),['Tot_latness','fahrzeug']]
out.loc[t_last].update(Data_t.groupby('fahrzeug').last()['Tot_latness'])
for t in range(startTime+dt,endTime+num_next_time*nex_time_dt*dt,dt):
Data_t = Data_d.loc[(Data_d['ist_ab_nach'] > t_last) &(Data_d['ist_ab_nach'] <= t),['Tot_latness','fahrzeug']]
if Data_t['fahrzeug'].size != 0:
out.loc[t,:].update(Data_t.groupby('fahrzeug').last()['Tot_latness'])
## out.loc[t,:] = out.loc[t_last,:]
# else:
# out.loc[t,:] = out.loc[t_last,:]
# out.loc[t,Data_t['fahrzeug'].unique()] = Data_t.groupby('fahrzeug').last()
t_last = t
out = out.dropna(axis='columns',how='all')
out = out.apply(smoothing, axis=0)
return_array = np.empty(((int((endTime-startTime)/dt)),num_next_time,out.shape[1]))
for w in range(int((endTime-startTime)/dt)):
return_array[w,:,:] = out.iloc[w:(num_next_time+w),:].to_numpy()
out_fin = np.append(out_fin,return_array,)
# return return_array
del(Data_t,Data_d,out,return_array)
pool = mp.Pool(ncoeur)
#out = pool.map(get_out,splitData)#np.array(,np.float32)
pool.close()
pool.join() # 28
out_fin = np.array(out_fin)
# In[ ]:
out = np.expand_dims(out,axis = 5)
out = np.expand_dims(out,axis = 6)
# In[ ]:
#with open('FrameT.pickle', 'wb') as f:
# pickle.dump(FrameT, f)
with open(os.getcwd()+'/DataPickles/out.pickle', 'wb') as o:
pickle.dump(out, o)
# ## LSTM model input_shape = (before_dt,FrameT.shape[3],FrameT.shape[4],FrameT.shape[5])),
model = Sequential([
BatchNormalization(name = 'batch_norm_0', input_shape = (before_dt,FrameT.shape[3],FrameT.shape[4],FrameT.shape[5])),
ConvLSTM2D(name ='conv_lstm_1',filters=32, kernel_size=(10, 1)
, padding='same', return_sequences=True,data_format ="channels_first"),
Dropout(0.2),
BatchNormalization(),
ConvLSTM2D(name ='conv_lstm_2',filters=16, kernel_size=(5, 1),data_format ="channels_first"
,padding='same', return_sequences=False),
Dropout(0.2),
BatchNormalization(),
Flatten(),
Dense(out.shape[3]),
RepeatVector(num_next_time),
Reshape((num_next_time,out.shape[3],1,1)),
#
ConvLSTM2D(name ='conv_lstm_3',filters=16, kernel_size=(10, 1)
, padding='same', return_sequences=True),
Dropout(0.1),
BatchNormalization(),
ConvLSTM2D(name ='conv_lstm_4',
filters = 32, kernel_size = (5, 1),
padding='same',
return_sequences = True),
# Reshape((10,202472)) ,
#
# TimeDistributed(Dense(out.shape[2]*20)),
##
# TimeDistributed(Dense(32,)),
#
TimeDistributed(Dense(1,activation = "sigmoid"), name='test'),
])
optimizer = optimizers.Adam(clipvalue=0.5)
#optimizer = optimizers.Adam(clipnorm=1.)
model.compile(loss='mse', optimizer = optimizer)
model.summary()
plot_model(model)
# In[ ]:
#batch_size = 60
#epochs = 10
#t1 = time.time()
##for day in range(len(Data['datum_nach'].unique())):
##def model_fit(day):
#error = np.array([])
#for e in range(epochs):
# for day in range(len(Data['datum_nach'].unique())-7):
# print('day :',day+1,'epoque :',e+1)
# model.fit(FrameT[day,:,:,:,:,:], out[day,:,:,:,:,:], epochs=1, batch_size=batch_size, verbose=1, shuffle=False)
# model.reset_states()
# for p in range(1,7):
# predicty = model.predict(FrameT[-p,:,:,:,:,:])
# errorp += np.sum(abs(predicty-out[-p,:,:,:,:,:]))
#
# print('Epoch',e,'Error:',errorp)
#
# error = np.append(error,errorp)
# errorp = 0
#
##pool = mp.Pool(ncoeur)
##
###out = pool.map(model_fit,)
##pool.close()
##pool.join() # 28
#
#print('time spend:',time.time()-1)
# In[ ]:
#hidden_layers = keras.backend.function(
#[model.layers[0].input], # we will feed the function with the input of the first layer
#[model.layers[0].output,] # we want to get the output of the second layer
#)
#print(np.asarray(hidden_layers(FrameT[0,:,:,:,:,:])))
# In[ ]:
#
#t1 = time.time()
#latness_predict = model.predict(FrameT[4:5,:,:,:,:])
#print('time spend:',t1-time.time())
#
## In[ ]:
#
#
#loss = model.history['loss']
#loss_val = model.history['val_loss']
#epochs_plot = range(epochs)
#plt.figure(2)
#plt.plot(epochs_plot, loss, 'bo', label='Training loss') #val_loss
#plt.plot(epochs_plot, loss_val, 'bs', label='Validation loss') #val_loss
#
#plt.title('Loss')
#plt.xlabel('Epochs')
#plt.ylabel('Loss')
#plt.legend()
#plt.show()
#plt.savefig('test.png')
#
#
## In[ ]:
#
#
#plt.figure(3,figsize = (12,5))
#
#which = 4
#
##plt.plot(predicty[:,:,which,0,0], label='Predict') #val_loss
#plt.plot(out[4,:.,:,which,0,0], label='Real') #val_loss/
#
#plt.title('Loss')
#plt.xlabel('Time')
#plt.ylabel('Latness')
#plt.legend()
#plt.show()
#out.shape
# In[ ]:
# model = Sequential()
# model.add(BatchNormalization(name = 'batch_norm_0', input_shape = (input_timesteps, num_links, 1, 1)))
# model.add(ConvLSTM2D(name ='conv_lstm_1',
# filters = 64, kernel_size = (10, 1),
# padding = 'same',
# return_sequences = True))
# model.add(Dropout(0.2, name = 'dropout_1'))
# model.add(BatchNormalization(name = 'batch_norm_1'))
# model.add(ConvLSTM2D(name ='conv_lstm_2',
# filters = 64, kernel_size = (5, 1),
# padding='same',
# return_sequences = False))
# model.add(Dropout(0.1, name = 'dropout_2'))
# model.add(BatchNormalization(name = 'batch_norm_2'))
# model.add(Flatten())
# model.add(RepeatVector(output_timesteps))
# model.add(Reshape((output_timesteps, num_links, 1, 64)))
# model.add(ConvLSTM2D(name ='conv_lstm_3',
# filters = 64, kernel_size = (10, 1),
# padding='same',
# return_sequences = True))
# model.add(Dropout(0.1, name = 'dropout_3'))
# model.add(BatchNormalization(name = 'batch_norm_3'))
# model.add(ConvLSTM2D(name ='conv_lstm_4',
# filters = 64, kernel_size = (5, 1),
# padding='same',
# return_sequences = True))
# model.add(TimeDistributed(Dense(units=1, name = 'dense_1', activation = 'relu')))
# #model.add(Dense(units=1, name = 'dense_2'))
# model
# In[ ]: