forked from wepe/TianChi-Industry4.0-Competition
-
Notifications
You must be signed in to change notification settings - Fork 0
/
feature_extract.py
331 lines (283 loc) · 14.5 KB
/
feature_extract.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
#coding=utf-8
import numpy as np
import pandas as pd
import time
import os
from multiprocessing import Pool
os.mkdir('data/feature')
############ draft表 类别特征编码 ##############
######################################################
draft_train = pd.read_csv('data/draft_data_train.csv')
draft_test = pd.read_csv('data/draft_data_test_new.csv')
draft_test['key_index'] = -999
draft_train_test = pd.concat([draft_train,draft_test],axis=0)
category_var = ['draft_param1','draft_param2','draft_param3','draft_param10','draft_param11','draft_param4','draft_param6','draft_param7']
for var in category_var:
var_dummies = pd.get_dummies(draft_train_test[var])
var_dummies.columns = [var+'_'+str(i) for i in range(var_dummies.shape[1])]
if var not in ['draft_param4','draft_param6','draft_param7']:
draft_train_test.drop(var,axis=1,inplace=True)
draft_train_test = pd.concat([draft_train_test,var_dummies],axis=1)
draft_train = draft_train_test[draft_train_test.key_index!=-999]
draft_test = draft_train_test[draft_train_test.key_index==-999]
draft_test.drop('key_index',axis=1,inplace=True)
draft_train.to_csv('data/feature/draft_train.csv',index=None)
draft_test.to_csv('data/feature/draft_test.csv',index=None)
############ param表 类别特征编码 ##############
######################################################
param_train = pd.read_csv('data/param_data_train.csv')
param_test = pd.read_csv('data/param_data_test_new.csv')
param_test['key_index'] = -999
param_train_test = pd.concat([param_train,param_test],axis=0)
category_var = ['param3','param4','param7','param8','param5','param9']
for var in category_var:
var_dummies = pd.get_dummies(param_train_test[var])
var_dummies.columns = [var+'_'+str(i) for i in range(var_dummies.shape[1])]
if var not in ['param5','param9']:
param_train_test.drop(var,axis=1,inplace=True)
param_train_test = pd.concat([param_train_test,var_dummies],axis=1)
############ param表 数值型特征 ##############
######################################################
"""
param5 4,6
param9 3.6, 3.1, 3.
对'param5','param9'也进行了编码,但保留原始字段
"""
param_train = param_train_test[param_train_test.key_index!=-999]
param_test = param_train_test[param_train_test.key_index==-999]
param_test.drop('key_index',axis=1,inplace=True)
param_train.to_csv('data/feature/param_train.csv',index=None)
param_test.to_csv('data/feature/param_test.csv',index=None)
############ timevarying 表 ############
######################################################
"""
test数据都有时序参数记录
train数据14675个product只有6940个有时序参数记录
提取特征:对每个product,提取各个参数的最小,最大,中值,平均值,众数,方差,记录个数(分tparam1=50之前的记录和全部记录，更加细致地分10个时间段)
"""
params = ['tparam1', 'tparam18', 'tparam14', 'tparam2', 'tparam10', 'tparam7','tparam9', 'tparam3', 'tparam8', 'tparam11', 'tparam17', 'tparam4','tparam5', 'tparam6', 'tparam16', 'tparam15', 'tparam12', 'tparam13']
sts = ['_min','_max','_median','_mean','_std','_cnt','_max_min']
######## testset ##########
##############################
tv_test = pd.read_csv('data/timevarying_param_test_new.csv',header=None)
tv_test.columns = ['product_no','param_name','param_value','add_time']
products = list(tv_test.product_no.unique())
#全部记录
tv_features = pd.DataFrame(columns=['product_no']+[p+s for s in sts for p in params])
def test_(product):
d = {'product_no':product}
tv_product = tv_test[tv_test.product_no==product]
for param in params:
tv_product_param = tv_product[tv_product.param_name==param]
d[param+'_min'] = tv_product_param.param_value.min()
d[param+'_max'] = tv_product_param.param_value.max()
d[param+'_median'] = tv_product_param.param_value.median()
d[param+'_mean'] = tv_product_param.param_value.mean()
d[param+'_std'] = tv_product_param.param_value.std()
d[param+'_cnt'] = tv_product_param.param_value.count()
d[param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
rst.append(pool.apply_async(test_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features = pd.concat([tv_features,i],axis=0)
tv_features.to_csv('data/feature/tv_features_test.csv',index=None)
#进度50%之前的记录
tv_features_half = pd.DataFrame(columns=['product_no']+[p+s for s in sts for p in params])
def test_(product):
d = {'product_no':product}
tv_product = tv_test[tv_test.product_no==product]
tv_product.add_time = tv_product.add_time.apply(lambda x:int(time.mktime(time.strptime(x,"%Y-%m-%d %H:%M:%S"))))
half_time = int(tv_product[(tv_product.param_name=='tparam1')&(tv_product.param_value==50.0)].add_time)
tv_product = tv_product[tv_product.add_time<half_time]
if tv_product.shape[0]==0:
print product
for param in params:
tv_product_param = tv_product[tv_product.param_name==param]
d[param+'_min'] = tv_product_param.param_value.min()
d[param+'_max'] = tv_product_param.param_value.max()
d[param+'_median'] = tv_product_param.param_value.median()
d[param+'_mean'] = tv_product_param.param_value.mean()
d[param+'_std'] = tv_product_param.param_value.std()
d[param+'_cnt'] = tv_product_param.param_value.count()
d[param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
print product
rst.append(pool.apply_async(test_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features_half = pd.concat([ tv_features_half,i],axis=0)
tv_features_half.to_csv('data/feature/tv_features_half_test.csv',index=None)
######## trainset ##########
##############################
tv_train = pd.read_csv('data/timevarying_param_train.csv',header=None)
tv_train.columns = ['product_no','key_index','param_name','param_value','add_time']
products = list(tv_train.product_no.unique())
#全部记录
tv_features = pd.DataFrame(columns=['product_no']+[p+s for s in sts for p in params])
def train_(product):
d = {'product_no':product}
tv_product = tv_train[tv_train.product_no==product]
for param in params:
tv_product_param = tv_product[tv_product.param_name==param]
d[param+'_min'] = tv_product_param.param_value.min()
d[param+'_max'] = tv_product_param.param_value.max()
d[param+'_median'] = tv_product_param.param_value.median()
d[param+'_mean'] = tv_product_param.param_value.mean()
d[param+'_std'] = tv_product_param.param_value.std()
d[param+'_cnt'] = tv_product_param.param_value.count()
d[param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
print product
rst.append(pool.apply_async(train_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features = pd.concat([tv_features,i],axis=0)
tv_features.to_csv('data/feature/tv_features_train.csv',index=None)
#进度50%之前的记录
tv_features_half = pd.DataFrame(columns=['product_no']+[p+s for s in sts for p in params])
def train_(product):
d = {'product_no':product}
tv_product = tv_train[tv_train.product_no==product]
tv_product.add_time = tv_product.add_time.apply(lambda x:int(time.mktime(time.strptime(x,"%Y-%m-%d %H:%M:%S"))))
half_time = int(tv_product[(tv_product.param_name=='tparam1')&(tv_product.param_value==50.0)].add_time)
tv_product = tv_product[tv_product.add_time<half_time]
for param in params:
tv_product_param = tv_product[tv_product.param_name==param]
d[param+'_min'] = tv_product_param.param_value.min()
d[param+'_max'] = tv_product_param.param_value.max()
d[param+'_median'] = tv_product_param.param_value.median()
d[param+'_mean'] = tv_product_param.param_value.mean()
d[param+'_std'] = tv_product_param.param_value.std()
d[param+'_cnt'] = tv_product_param.param_value.count()
d[param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
rst.append(pool.apply_async(train_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features_half = pd.concat([ tv_features_half,i],axis=0)
tv_features_half.to_csv('data/feature/tv_features_half_train.csv',index=None)
################# 分10个时间段统计 #################
######################################################
stage = ['stg1_','stg2_','stg3_','stg4_','stg5_','stg6_','stg7_','stg8_','stg9_','stg10_']
params = ['tparam1', 'tparam18', 'tparam14', 'tparam2', 'tparam10', 'tparam7','tparam9', 'tparam3', 'tparam8', 'tparam11', 'tparam17', 'tparam4','tparam5', 'tparam6', 'tparam16', 'tparam15', 'tparam12', 'tparam13']
sts = ['_min','_max','_median','_mean','_std','_cnt','_max_min']
######## testset ##########
##############################
tv_test = pd.read_csv('data/timevarying_param_test_new.csv',header=None)
tv_test.columns = ['product_no','param_name','param_value','add_time']
products = list(tv_test.product_no.unique())
tv_features = pd.DataFrame(columns=['product_no']+[stg+p+s for stg in stage for s in sts for p in params])
def test_(product):
d = {'product_no':product}
tv_product = tv_test[tv_test.product_no==product]
tv_product.add_time = tv_product.add_time.apply(lambda x:int(time.mktime(time.strptime(x,"%Y-%m-%d %H:%M:%S"))))
t = tv_product.add_time.describe(percentiles=[0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9])
split_point = [0,int(t['10%']),int(t['20%']),int(t['30%']),int(t['40%']),int(t['50%']),int(t['60%']),int(t['70%']),int(t['80%']),int(t['90%']),1e11]
for i in range(10):
stg = stage[i]
tv_product_stg = tv_product[(split_point[i]<tv_product.add_time)&(tv_product.add_time<split_point[i+1])]
for param in params:
tv_product_param = tv_product_stg[tv_product_stg.param_name==param]
d[stg+param+'_min'] = tv_product_param.param_value.min()
d[stg+param+'_max'] = tv_product_param.param_value.max()
d[stg+param+'_median'] = tv_product_param.param_value.median()
d[stg+param+'_mean'] = tv_product_param.param_value.mean()
d[stg+param+'_std'] = tv_product_param.param_value.std()
d[stg+param+'_cnt'] = tv_product_param.param_value.count()
d[stg+param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
print product
rst.append(pool.apply_async(test_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features = pd.concat([tv_features,i],axis=0)
tv_features.to_csv('data/feature/tv_stage10_test.csv',index=None)
######## trainset ##########
##############################
tv_train = pd.read_csv('data/timevarying_param_train.csv',header=None)
tv_train.columns = ['product_no','key_index','param_name','param_value','add_time']
products = list(tv_train.product_no.unique())
tv_features = pd.DataFrame(columns=['product_no']+[stg+p+s for stg in stage for s in sts for p in params])
def train_(product):
d = {'product_no':product}
tv_product = tv_train[tv_train.product_no==product]
tv_product.add_time = tv_product.add_time.apply(lambda x:int(time.mktime(time.strptime(x,"%Y-%m-%d %H:%M:%S"))))
t = tv_product.add_time.describe(percentiles=[0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9])
split_point = [0,int(t['10%']),int(t['20%']),int(t['30%']),int(t['40%']),int(t['50%']),int(t['60%']),int(t['70%']),int(t['80%']),int(t['90%']),1e11]
for i in range(10):
stg = stage[i]
tv_product_stg = tv_product[(split_point[i]<tv_product.add_time)&(tv_product.add_time<split_point[i+1])]
for param in params:
tv_product_param = tv_product_stg[tv_product_stg.param_name==param]
d[stg+param+'_min'] = tv_product_param.param_value.min()
d[stg+param+'_max'] = tv_product_param.param_value.max()
d[stg+param+'_median'] = tv_product_param.param_value.median()
d[stg+param+'_mean'] = tv_product_param.param_value.mean()
d[stg+param+'_std'] = tv_product_param.param_value.std()
d[stg+param+'_cnt'] = tv_product_param.param_value.count()
d[stg+param+'_max_min'] = tv_product_param.param_value.max() - tv_product_param.param_value.min()
this_tv_features = pd.DataFrame(d,index=[0])
return this_tv_features
rst = []
pool = Pool(12)
for product in products:
print product
rst.append(pool.apply_async(train_,args=(product,)))
pool.close()
pool.join()
rst = [i.get() for i in rst]
for i in rst:
tv_features = pd.concat([tv_features,i],axis=0)
tv_features.to_csv('data/feature/tv_stage10_train.csv',index=None)
##################### totaltime #####################
######################################################
#for each product,get its max(add_time) - min(add_time)
tv_train = pd.read_csv('data/timevarying_param_train.csv',header=None)
tv_train.columns = ['product_no','key_index','param_name','param_value','add_time']
tv_train.drop('key_index',axis=1,inplace=True)
tv_test = pd.read_csv('data/timevarying_param_test_new.csv',header=None)
tv_test.columns = ['product_no','param_name','param_value','add_time']
tv = pd.concat([tv_train,tv_test],axis=0)
products = list(tv.product_no.unique())
def bar(product):
this_product = tv[tv.product_no==product]
this_product.add_time = this_product.add_time.apply(lambda x:int(time.mktime(time.strptime(x,"%Y-%m-%d %H:%M:%S"))))
total_time = this_product.add_time.max() - this_product.add_time.min()
return product,total_time
product_totaltime = Pool(12).map(bar,products)
product = np.array([i[0] for i in product_totaltime])
totaltime = np.array([i[1] for i in product_totaltime])
product_totaltime = pd.DataFrame(product,columns=['product_no'])
product_totaltime['totaltime'] = totaltime
product_totaltime.to_csv('data/feature/continue_time.csv',index=None)