-
Notifications
You must be signed in to change notification settings - Fork 5
/
generator.py
310 lines (282 loc) · 11.4 KB
/
generator.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
import os, sys
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from collections import OrderedDict
from collections import Counter
from operator import itemgetter
# Hawkes model is from https://omitakahiro.github.io/Hawkes/index.html
from modules import Hawkes as hk
para = {'mu':0.1, 'alpha':0.3, 'beta':0.6}
mu_t = lambda x: (1.0 + 0.8*np.sin(2*np.pi*x/100)) * 0.2 # baseline function for overlay
itv = [0,360000]
demo_itv = [0,360]
np.random.seed(42)
downsampling = {'Trump': 10}
#downsampling = {'taxi': 20, 'Trump': 20}
def downsampling_dataset(timestamps, dataset_name):
print('Down-sampling', dataset_name, 'dataset by', downsampling[dataset_name])
return timestamps[::downsampling[dataset_name]]
def purge_duplicate_events(timestamps, types):
timestamps = timestamps.tolist()
types = types.tolist()
del_indices = []
events = [(ts, ty) for ts, ty in zip(timestamps, types)]
events_next = events[1:]
np.where([e[0]==en[0] and e[1]==en[1] for e, en in zip(events[:-1], events_next)])
for i in range(1, len(timestamps)):
if timestamps[i]==timestamps[i-1] and types[i]==types[i-1]:
del_indices.append(i)
for ind in sorted(del_indices, reverse=True):
del timestamps[ind]
del types[ind]
return timestamps, types
def keep_top_k_types(types, keep_classes=10):
types_counter = OrderedDict(sorted(Counter(types).items(), key=itemgetter(1), reverse=True))
type2supertype = OrderedDict()
for i, (type_, _) in enumerate(types_counter.items()):
if i > keep_classes:
type2supertype[type_] = keep_classes + 1
else:
type2supertype[type_] = i + 1
types_new = [type2supertype[ty] for ty in types]
return np.array(types_new)
def hawkes_demo():
hk_model = hk.simulator().set_kernel('exp').set_baseline('const').set_parameter(para)
T = hk_model.simulate(demo_itv)
hk_model.plot_l()
plt.savefig('hawkes_intensity.png')
plt.close()
hk_model.plot_N()
plt.savefig('hawkes_event_counts.png')
plt.close()
def sin_hawkes_overlay_demo():
hk_model = hk.simulator().set_kernel('exp').set_baseline('custom',l_custom=mu_t).set_parameter(para)
T = hk_model.simulate(demo_itv)
hk_model.plot_l()
plt.savefig('sin_hawkes_overlay_intensity.png')
plt.close()
hk_model.plot_N()
plt.savefig('sin_hawkes_overlay_event_counts.png')
plt.close()
def create_sin_data():
omega = 1.0
points = 10000
num_marks = 7
x = np.linspace(0, points, 3*points)
y_ = 10*np.sin(omega*x)
y = y_ + 11
gaps=y
timestamp = np.cumsum(gaps)
types = []
if y_[0]<y_[1]:
types.append(1)
for i in range(len(y_[1:])):
#if y_[i]>=0.:
# types.append(0)
#else:
# types.append(1)
if y_[i]>=0. and y_[i]>y_[i-1]:
types.append(1)
if y_[i]>=0. and y_[i]<y_[i-1]:
types.append(2)
if y_[i]<0. and y_[i]<y_[i-1]:
types.append(3)
if y_[i]<0. and y_[i]>y_[i-1]:
types.append(4)
types = np.array(types)
plt.plot(x[:25], y[:25], 'o', color='black');
plt.savefig('data/sin.png')
plt.close()
return gaps, timestamp, types
def create_hawkes_data():
hawkes_demo()
hk_model = hk.simulator().set_kernel('exp').set_baseline('const').set_parameter(para)
timestamp = hk_model.simulate([0, 360000])
gaps = timestamp[1:] - timestamp[:-1]
return gaps, timestamp
def create_sin_hawkes_overlay_data():
sin_hawkes_overlay_demo()
hk_model = hk.simulator().set_kernel('exp').set_baseline('custom',l_custom=mu_t).set_parameter(para)
timestamp = hk_model.simulate([0, 360000])
gaps = timestamp[1:] - timestamp[:-1]
return gaps, timestamp
def create_taxi_data():
# https://s3.amazonaws.com/nyc-tlc/trip+data/yellow_tripdata_2019-01.csv
# https://s3.amazonaws.com/nyc-tlc/trip+data/yellow_tripdata_2019-02.csv
taxi_df_jan = pd.read_csv(
'./data/yellow_tripdata_2019-01.csv',
usecols=["tpep_pickup_datetime", "PULocationID", "DOLocationID"])
taxi_df_feb = pd.read_csv(
'./data/yellow_tripdata_2019-02.csv',
usecols=["tpep_pickup_datetime", "PULocationID", "DOLocationID"])
taxi_df = taxi_df_jan.append(taxi_df_feb)
taxi_df = taxi_df[taxi_df.PULocationID == 237]
taxi_df['tpep_pickup_datetime'] = pd.to_datetime(taxi_df['tpep_pickup_datetime'], errors='coerce')
taxi_df = taxi_df[(taxi_df['tpep_pickup_datetime'].dt.year == 2019)]
taxi_df = taxi_df[(taxi_df['tpep_pickup_datetime'].dt.month < 3)]
taxi_df = taxi_df.sort_values('tpep_pickup_datetime')
taxi_types = taxi_df['DOLocationID'].values
#taxi_timestamps = taxi_timestamps.sort_values().astype(np.int64)
taxi_timestamps = pd.DatetimeIndex(taxi_df['tpep_pickup_datetime']).astype(np.int64)/1000000000
taxi_timestamps = np.array(taxi_timestamps)
taxi_timestamps -= taxi_timestamps[0]
taxi_timestamps = taxi_timestamps[:-1]
taxi_types = taxi_types[:-1]
taxi_types = keep_top_k_types(taxi_types)
dataset_name = 'taxi'
if dataset_name in downsampling:
taxi_timestamps = downsampling_dataset(taxi_timestamps, dataset_name)
taxi_types = downsampling_dataset(taxi_types, dataset_name)
taxi_gaps = taxi_timestamps[1:] - taxi_timestamps[:-1]
plt.plot(taxi_gaps[:100])
plt.ylabel('Gaps')
plt.savefig('data/taxi_gaps.png')
plt.close()
return taxi_gaps, taxi_timestamps, taxi_types
def create_911_traffic_data():
call_df = pd.read_csv('./data/911.csv')
call_df = call_df[call_df['zip'].isnull()==False] # Ignore calls with NaN zip codes
print('Types of Emergencies')
print(call_df.title.apply(lambda x: x.split(':')[0]).value_counts())
call_df['type'] = call_df.title.apply(lambda x: x.split(':')[0])
print('Subtypes')
for each in call_df.type.unique():
subtype_count = call_df[call_df.title.apply(lambda x: x.split(':')[0]==each)].title.value_counts()
print('For', each, 'type of Emergency, we have ', subtype_count.count(), 'subtypes')
print(subtype_count[subtype_count>100])
print('Out of 3 types taking Traffic type considering only Traffic')
call_data = call_df[call_df['type']=='Traffic']
call_data['timeStamp'] = pd.to_datetime(call_data['timeStamp'], errors='coerce')
print("We have timeline from", call_data['timeStamp'].min(), "to", call_data['timeStamp'].max())
call_data = call_data.sort_values('timeStamp')
call_timestamps = pd.DatetimeIndex(call_data['timeStamp']).astype(np.int64)/1000000000
#call_timestamps = call_data.sort_values().astype(np.int64)
call_timestamps = np.array(call_timestamps)
call_timestamps -= call_timestamps[0]
call_types = call_data['zip'].values
call_types = keep_top_k_types(call_types)
dataset_name = 'call'
if dataset_name in downsampling:
call_timestamps = downsampling_dataset(call_timestamps, dataset_name)
call_types = downsampling_dataset(call_types, dataset_name)
call_gaps = call_timestamps[1:] - call_timestamps[:-1]
plt.plot(call_gaps[:100])
plt.ylabel('Gaps')
plt.xlabel('timeline')
plt.savefig('data/call_traffic_gaps.png')
plt.close()
return call_gaps, call_timestamps, call_types
def create_911_ems_data():
call_df = pd.read_csv('./data/911.csv')
call_df = call_df[call_df['zip'].isnull()==False] # Ignore calls with NaN zip codes
call_df['type'] = call_df.title.apply(lambda x: x.split(':')[0])
print('Out of 3 types taking EMS type considering only EMS')
call_data = call_df[call_df['type']=='EMS']
call_data['timeStamp'] = pd.to_datetime(call_data['timeStamp'], errors='coerce')
print("We have timeline from", call_data['timeStamp'].min(), "to", call_data['timeStamp'].max())
call_data = call_data.sort_values('timeStamp')
call_timestamps = pd.DatetimeIndex(call_data['timeStamp']).astype(np.int64)/1000000000
#call_timestamps = call_data.sort_values().astype(np.int64)
call_timestamps = np.array(call_timestamps)
call_timestamps -= call_timestamps[0]
call_types = call_data['zip'].values
call_types = keep_top_k_types(call_types)
dataset_name = 'call'
if dataset_name in downsampling:
call_timestamps = downsampling_dataset(call_timestamps, dataset_name)
call_types = downsampling_dataset(call_types, dataset_name)
call_gaps = call_timestamps[1:] - call_timestamps[:-1]
plt.plot(call_gaps[:100])
plt.ylabel('Gaps')
plt.xlabel('timeline')
plt.savefig('data/call_ems_gaps.png')
plt.close()
return call_gaps, call_timestamps, call_types
def generate_dataset():
os.makedirs('./data', exist_ok=True)
#os.chdir('./data')
if not os.path.isfile("sin.txt"):
print('Generating sin data')
gaps, timestamps, types = create_sin_data()
timestamps, types = purge_duplicate_events(timestamps, types)
np.savetxt('data/sin.txt', timestamps)
np.savetxt('data/sin_types.txt', types)
# if not os.path.isfile("hawkes.txt"):
# print('Generating hawkes data')
# gaps, timestamps = create_hawkes_data()
# timestamps, types = purge_duplicate_events(timestamps, types)
# np.savetxt('hawkes.txt', timestamps)
# if not os.path.isfile("sin_hawkes_overlay.txt"):
# print('Generating sin_hawkes_overlay data')
# gaps, timestamps = create_sin_hawkes_overlay_data()
# timestamps, types = purge_duplicate_events(timestamps, types)
# np.savetxt('sin_hawkes_overlay.txt', timestamps)
if not os.path.isfile("911_traffic.txt"):
print('Generating 911 data')
gaps, timestamps, types = create_911_traffic_data()
timestamps, types = purge_duplicate_events(timestamps, types)
np.savetxt('data/911_traffic.txt', timestamps)
np.savetxt('data/911_traffic_types.txt', types)
if not os.path.isfile("911_ems.txt"):
print('Generating 911 data')
gaps, timestamps, types = create_911_ems_data()
timestamps, types = purge_duplicate_events(timestamps, types)
np.savetxt('data/911_ems.txt', timestamps)
np.savetxt('data/911_ems_types.txt', types)
if not os.path.isfile("taxi.txt"):
print('Generating taxi data')
gaps, timestamps, types = create_taxi_data()
timestamps = np.array(timestamps).astype(np.float32)
types = np.array(types).astype(np.float32)
timestamps, types = purge_duplicate_events(timestamps, types)
np.savetxt('data/taxi.txt', timestamps)
np.savetxt('data/taxi_types.txt', types)
#os.chdir('../')
def create_twitter_data(dataset_name, keep_classes=10):
delimiter=' '
if dataset_name in ['Movie', 'Delhi', 'Verdict', 'Fight']:
delimiter='\t'
twitter_df = pd.read_csv('./data/'+dataset_name+'.txt', delimiter=delimiter, header=None)
twitter_df = twitter_df.values[::-1]
#twitter_df = twitter_df[1]
timestamps = twitter_df[:, 1]
timestamps -= timestamps[0]
gaps = timestamps[1:] - timestamps[:-1]
types = twitter_df[:, 0]
types_counter = OrderedDict(sorted(Counter(types).items(), key=itemgetter(1), reverse=True))
type2supertype = OrderedDict()
for i, (type_, _) in enumerate(types_counter.items()):
if i > keep_classes:
type2supertype[type_] = keep_classes + 1
else:
type2supertype[type_] = i + 1
types_new = [type2supertype[ty] for ty in types]
types = types_new
if dataset_name in downsampling:
plt.plot(gaps)
plt.ylabel('all_Gaps_before_downsample')
plt.savefig(dataset_name+'_all_gaps_before_downsample.png')
plt.close()
timestamps = downsampling_dataset(timestamps, dataset_name)
types = downsampling_dataset(types, dataset_name)
plt.plot(gaps[:100])
plt.ylabel('Gaps')
plt.savefig(dataset_name+'_gaps.png')
plt.close()
plt.plot(gaps)
plt.ylabel('all_Gaps')
plt.savefig(dataset_name+'_all_gaps.png')
plt.close()
return gaps, timestamps, types
def generate_twitter_dataset(twitter_dataset_names):
os.makedirs('./data', exist_ok=True)
#os.chdir('./data')
for dataset_name in twitter_dataset_names:
if not os.path.isfile(dataset_name+'.txt'):
print('Generating', dataset_name, 'data')
gaps, timestamps, types = create_twitter_data(dataset_name)
timestamps, types = purge_duplicate_events(np.array(timestamps), np.array(types))
np.savetxt(dataset_name+'.txt', timestamps)
np.savetxt(dataset_name+'_types.txt', types)
#os.chdir('../')