-
Notifications
You must be signed in to change notification settings - Fork 2
/
xgb_whole.py
351 lines (251 loc) · 11.9 KB
/
xgb_whole.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
# -*- coding: utf-8 -*-
"""
@author: Gene Baratheon
@Email : [email protected]
@Main : XGBOOST单模型：全局统计
"""
import pandas as pd
import numpy as np
import time
from sklearn.metrics import log_loss
from collections import Counter
import extract_feature_whole as ext_feat_wh
import lightgbm as lgb
import warnings
from sklearn.preprocessing import LabelEncoder,OneHotEncoder
from scipy import sparse
from sklearn.linear_model import LogisticRegression
import xgboost as xgb
warnings.filterwarnings("ignore")
#线下xgboost
def xgbCV(X_train,X_test):
Y_train=X_train.pop('is_trade')
Y_test=X_test.pop('is_trade')
index_train=X_train.pop('instance_id')
index_test =X_test.pop('instance_id')
print('train DMatrix')
train = xgb.DMatrix(X_train,label=Y_train)
del X_train ; gc.collect()
print('test DMatrix')
test = xgb.DMatrix(X_test,label=Y_test)
del X_test ; gc.collect()
print('xgb model start......')
params={'booster':'gbtree',
'objective':'binary:logistic',
'eval_metric':'logloss',
#'gamma':0.1,
'min_child_weight':1.1,
'max_depth':5,
'lambda':10,
'subsample':0.75,
'colsample_bytree':0.7,
'colsample_bylevel':0.7,
'eta': 0.03,
'tree_method':'hist',
'seed':0,
'nthread':10
}
watchlist = [(train,'train'),(test,'val')]
model = xgb.train(params,train,num_boost_round=5000,evals=watchlist,early_stopping_rounds=400)
#predict test set
Pre_test=model.predict(test)
feat_imp=model.get_fscore()
loss = log_loss(Y_test,Pre_test)
print('xgb logloss is: ',loss)
best_int=model.best_iteration
# #保存线下结果
# sub = pd.DataFrame()
# sub['instance_id'] = list(index_test)
# sub['instance_id'] = list(index_test)
# sub['predicted_score'] = list(Pre_test)
# sub.to_csv('../result/20180423_gene_xgb_offline.txt',sep=" ",index=False)
return feat_imp,Pre_test,loss,best_int
#线上xgboost
def sub(X_train,X_test,best_int):
Y_train=X_train.pop('is_trade')
Y_test=X_test.pop('is_trade')
index_train=X_train.pop('instance_id')
index_test =X_test.pop('instance_id')
#权重
train_weight = X_train.pop('weight')
print('train DMatrix')
train = xgb.DMatrix(X_train,label = Y_train,weight = train_weight)
del X_train ; gc.collect()
print('test DMatrix')
test = xgb.DMatrix(X_test)
del X_test ; gc.collect()
print('xgb model start......')
params={'booster':'gbtree',
'objective': 'binary:logistic',
'eval_metric':'logloss',
#'gamma':0.1,
'min_child_weight':1.1,
'max_depth':5,
'lambda':10,
'subsample':0.75,
'colsample_bytree':0.7,
'colsample_bylevel':0.7,
'eta': 0.03,
'tree_method':'hist',
'seed':0,
'nthread':10
}
watchlist = [(train,'train')]
model = xgb.train(params,train,num_boost_round=best_int,evals=watchlist)
#predict test set
Y_test=model.predict(test)
sub = pd.DataFrame()
sub['instance_id'] = list(index_test)
sub['instance_id'] = list(index_test)
sub['predicted_score'] = list(Y_test)
sub.to_csv('./20180515_gene_final.txt',sep=" ",index=False)
return model
#保存feature_importance
def save_feature_importance(feat_imp):
feat_imp=sorted(feat_imp.items(),key=lambda x:x[1],reverse=True)
features = [i[0] for i in feat_imp ]
importances = [i[1] for i in feat_imp ]
feat = pd.DataFrame()
feat['feature'] = features
feat['importance'] = importances
feat.to_csv('../feature/gene/feature_importance_xgb_gene.csv',sep=",",index=False)
#交叉验证
from sklearn.cross_validation import KFold
def CV_train(cv_data,n_folds=5):
logloss_sum=[]
cv_data=cv_data.reset_index()
kf=KFold(len(cv_data),n_folds=n_folds,shuffle=False)
for traincv,testcv in kf:
print('train part: ',traincv,' test part: ',testcv)
train=cv_data.loc[traincv,:];test=cv_data.loc[testcv,:]
_,Pre_test,loss=xgbCV(train,test)
logloss_sum.append(loss)
print('CV all logloss are: ',logloss_sum)
logloss_avg=np.mean(logloss_sum)
print('CV average logloss is : ',logloss_avg)
return logloss_avg
import gc
def change_dtypes(data):
all_data_int = data.select_dtypes(include = ['int64'])
all_data_convert_int = all_data_int.apply(pd.to_numeric, downcast = 'unsigned')
data[all_data_convert_int.columns] = all_data_convert_int
del all_data_int ; del all_data_convert_int
gc.collect()
return data
#保存添加属性
def save_add_feature(data,cols,path):
print('save add feature: ',cols)
t0 = data[cols]
print('save add shape: ',t0.shape)
t0.to_csv(path,index=False)
del t0 ; gc.collect()
if __name__ == "__main__":
start_time=time.time()
print('----------------------------读取基本特征-----------------------------------------------------------')
#1. 读取特征
data = pd.read_csv('../feature/gene/basic_all.csv')
print('data shape: ',data.shape) # 12161692
print('----------------------------数值特征贝叶斯平滑------------------------------------------------------')
#2. 暂无
print('----------------------------增加一些特征------------------------------------------------------------')
#3.增加5月6日特征: 18
feature_add1 = pd.read_csv('../feature/gene/add_0506.csv')
data = pd.merge(data,feature_add1,on='instance_id',how='left')
del feature_add1
gc.collect()
#3.增加5月7日特征: 5
feature_add2 = pd.read_csv('../feature/gene/add_0507.csv')
data = pd.merge(data,feature_add2,on='instance_id',how='left')
del feature_add2
gc.collect()
#4.增加5月8日特征: 3
feature_add3 = pd.read_csv('../feature/gene/add_0508.csv')
data = pd.merge(data,feature_add3,on='instance_id',how='left')
del feature_add3
gc.collect()
#4.增加5月9日特征: 4
feature_add4 = pd.read_csv('../feature/gene/add_0509.csv')
data = pd.merge(data,feature_add4,on='instance_id',how='left')
del feature_add4
gc.collect()
#5.增加前后15分钟点击: 15
print("load fast_pre_back_click_time_all_data_userid.....")
user_data1 = pd.read_csv("../training_data/round2_temp_data/sliding_15_all_data_userid1_b.csv", sep = ',')
user_data2 = pd.read_csv("../training_data/round2_temp_data/sliding_15_all_data_userid2_b.csv", sep = ',')
user_data = user_data1.append(user_data2, ignore_index = True)
del user_data1 ; del user_data2
gc.collect()
print("merge fast_pre_back_click_time_all_data_userid")
data = pd.merge(data, user_data, how = 'left', on = ['instance_id'])
del user_data
gc.collect()
data = data.fillna(0)
print('data shape after add: ',data.shape)
print('----------------------------删除一些特征------------------------------------------------------------')
#5.删除特征
drop_list=['context_timestamp','user_id','shop_id','item_id','property_0','property_1','item_category_list','item_brand_id']
data=data.drop(drop_list,axis=1) ; gc.collect()
print('data shape after drop: ',data.shape) # 320+45-8=357
print('----------------------------one_hot编码------------------------------------------------------------')
lr=LabelEncoder()
#1.ecoder的特征id
feat_set = ['item_city_id','user_occupation_id','user_gender_id',
'context_page_id','time_interval','week']
#4.进行encoder
for col in feat_set:
print('encoder feature is: ',col)
data[col]=lr.fit_transform(data[col])
#6.type transform
data = change_dtypes(data) ; gc.collect()
print('all data shape: ',data.shape)
print('----------------------------------线上-------------------------------------------------------------')
#1. 预测集
test_a = pd.read_csv('../training_data/round2_ijcai_18_test_a_20180425.txt',sep=" ") #519888
test_b = pd.read_csv('../training_data/round2_ijcai_18_test_b_20180510.txt',sep=" ") #1209768
testa_ins = list(test_a.instance_id)
testb_ins = list(test_b.instance_id)
del test_a ; del test_b
gc.collect()
test = data[data['instance_id'].isin(testb_ins)] #1209768
#2. 训练集
train_ins = list(data.instance_id)
train_ins = list(set(train_ins).difference(set(testa_ins)))
train_ins = list(set(train_ins).difference(set(testb_ins)))
train = data[data['instance_id'].isin(train_ins)] #得到训练集 10432036
train = train[train.day!=6] #去掉6号
del data ; gc.collect()
print('train shape: ',train.shape,' test shape: ',test.shape)
#3. 训练
best_int = 3300
train['weight'] = train.day.apply(lambda x:1 if x == 7 else 0.8) #增加权重
model = sub(train, test, best_int)
#计算算法时间
end_time=time.time()
print('all time is: %.1f'%(end_time-start_time),' 秒')
print('----------------------------------线下-------------------------------------------------------------')
#xgb CV 使用7号最后一个小时进行验证
#1. 得到测试集
test_offline = data[(data['day'] == 7) & (data['hour'] == 11)]
test_offline_ins = list(test_offline.instance_id)
#2.去掉线上
test_a = pd.read_csv('../training_data/round2_ijcai_18_test_a_20180425.txt',sep=" ") #519888
test_b = pd.read_csv('../training_data/round2_ijcai_18_test_b_20180510.txt',sep=" ") #1209768
testa_ins = list(test_a.instance_id)
testb_ins = list(test_b.instance_id)
del test_a ; del test_b
gc.collect()
train_ins = list(data.instance_id)
test_ins=test_offline_ins+testa_ins+testb_ins
train_ins = list(set(train_ins).difference(set(test_ins)))
#2. 得到训练集
train = data[data['instance_id'].isin(train_ins)]
del data ; gc.collect()
print('train shape: ',train.shape,' test shape: ',test_offline.shape)
#3. 线下训练
feat_imp,Pre_test,loss,best_int=xgbCV(train,test_offline)
#4. 保存feature_importance
print('best interation is: ',best_int)
save_feature_importance(feat_imp) ; gc.collect()
#计算算法时间
end_time=time.time()
print('all time is: %.1f'%(end_time-start_time),' 秒')