-
Notifications
You must be signed in to change notification settings - Fork 1
/
demoV2.py
260 lines (211 loc) · 8.79 KB
/
demoV2.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
# -*- coding:utf-8 -*-
import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import cross_validate
from sklearn.metrics import make_scorer
from sklearn.metrics import f1_score
#import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.metrics import recall_score
from sklearn.metrics import precision_score
from sklearn.metrics import f1_score
from sklearn.metrics import roc_auc_score
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.model_selection import GridSearchCV
import sys
sys.path.append('/Users/wanjun/Desktop/LightGBM/python-package')
import lightgbm as lgb
import gc
import mlxtend
from sklearn.metrics import confusion_matrix,roc_curve
from bayes_opt import BayesianOptimization
from sklearn.model_selection import StratifiedKFold
from matplotlib import pyplot as plt
import seaborn as sns
sns.set_style('whitegrid')
import warnings
warnings.filterwarnings("ignore")
#994731
#print(len(data))
#print(len(np.unique(data.index)))
#0 977884
#1 12122
#-1 4725
#pd.value_counts(data['label'])
#记分函数
def score_atc(y_true,y_pred_prob):
fpr, tpr, thresholds = roc_curve(y_true,y_pred_prob, pos_label=1)
score=0.4*tpr[(fpr>=0.001)][0]+0.3*tpr[(fpr>=0.005)][0]+0.3*tpr[(fpr>=0.01)][0]
return score
def train_test(x,y):
x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.1)
lgbc=lgb.LGBMClassifier(n_estimators=500,max_depth=-1,num_leaves=65,learning_rate=0.01,subsample=0.8,sub_feature=0.8,random_state=0,n_jobs=2,objective='binary',is_unbalance=True)
lgbc.fit(x_train,y_train)
pred=lgbc.predict(x_test)
pred_prob=lgbc.predict_proba(x_test)
print(precision_score(y_test,pred,pos_label=1))
print(recall_score(y_test,pred,pos_label=1))
print(roc_auc_score(y_test,pred_prob[:,1]))
return y_test,pred_prob
def lgbmodel(x,y):
#train_data = lgb.Dataset(data, label=label)
#class_weight={0:1,1:1.41}
lgbc=lgb.LGBMClassifier(n_estimators=500,max_depth=-1,num_leaves=80,learning_rate=0.01,subsample=0.8,random_state=0,n_jobs=4,objective='binary',is_unbalance=True)
scores=cross_validate(estimator=lgbc,X=x,y=y,cv=10,scoring=make_scorer(roc_auc_score),n_jobs=1,verbose=-1)
print(scores['test_score'].mean())
return scores
def lgbccv(n_estimators,num_leaves,min_child_samples,reg_alpha,reg_lambda,subsample,colsample_bytree):
skf=StratifiedKFold(n_splits=5)
skf.get_n_splits(data,label)
score=[]
for train_index, test_index in skf.split(data,label):
lgbc=lgb.LGBMClassifier(n_estimators=int(n_estimators),max_depth=-1,num_leaves=int(num_leaves),
learning_rate=0.01,subsample=max(min(subsample, 1), 0),
colsample_bytree=max(min(colsample_bytree, 1), 0),random_state=666,
min_child_samples=int(min_child_samples),
reg_alpha=max(reg_alpha,0),reg_lambda=max(reg_lambda,0),
n_jobs=4,objective='binary',
is_unbalance=True)
lgbc.fit(data[train_index],label[train_index])
pred_prob=lgbc.predict_proba(data[test_index])
score.append(score_atc(label[test_index],pred_prob[:,1]))
return np.array(score).mean()
def rfmodel(x,y):
pass
def select_feature():
pass
def sub_pred():
data=pd.read_csv('data/atec_anti_fraud_train.csv',index_col=0)
data.loc[data.label==-1,'label']=1
#data=data[data['label']!=-1]
data.fillna(-1,inplace=True)
y=data['label']
print(pd.value_counts(y))
x=data[data.columns[data.columns!='label']]
x.drop('date',axis=1,inplace=True)
#for i in range(1,20):
# if i==5:
# continue
# x['f{}'.format(i)]=x['f{}'.format(i)].astype('category')
gc.collect()
lgbc=lgb.LGBMClassifier(n_estimators=500,max_depth=-1,num_leaves=100,learning_rate=0.01,subsample=0.8,sub_feature=0.8,random_state=0,n_jobs=-1,objective='binary')
lgbc.fit(x,y)
test=pd.read_csv('data/atec_anti_fraud_test_b.csv',index_col=0)
test.fillna(-1,inplace=True)
test.drop('date',axis=1,inplace=True)
#for i in range(1,20):
# if i==5:
# continue
# test['f{}'.format(i)]=test['f{}'.format(i)].astype('category')
pred_prob=lgbc.predict_proba(test)
temp=pd.DataFrame(pred_prob[:,1],index=test.index)
temp.reset_index(inplace=True)
temp.columns=['id','score']
temp.to_csv('test_wtf_b.csv',index=None)
def compelet_data():
data=pd.read_csv('./data/atec_anti_fraud_train.csv',index_col=0)
y=data['label']
y[y==-1]=1
x=data[data.columns[data.columns!='label']]
gc.collect()
print('load data')
print('counting')
#x_complete=iterforest.IterImput().complete(x.values)
x_complete=simple.SimpleFill(fill_method='mean').complete(x.values)
print('count out')
x_complete=pd.DataFrame(x_complete)
print('df')
x_complete.to_csv('test_comp.csv')
'''
colsample_bytree：0.9547
min_child_samples 5.3656
n_estimators：1450.1096
num_leaves 119.3153 |
reg_alpha 0.1930 |
reg_lambda： 8.6361 |
subsample 0.8495 |
'''
def obsearch():
data=pd.read_csv('test_comp.csv',index_col=0)
data=data.values
label=pd.read_csv('label.csv',index_col=0,header=None)
label=label.values.reshape(-1)
print('load data')
init_points=10
num_iter=40
lgbBO = BayesianOptimization(lgbccv, {'n_estimators': (500,1500),
'num_leaves': (60, 120),
'min_child_samples': (5,100),
'reg_alpha': (0,10),
'reg_lambda': (0,10),
'subsample':(0.5,1.0),
'colsample_bytree':(0.5,1.0)
})
lgbBO.maximize(init_points=init_points, n_iter=num_iter)
print(lgbBO.res['max']['max_val'])
def statis_data():
#每个id唯一
data=pd.read_csv('data/atec_anti_fraud_train.csv',index_col=0)
data.sort_values('date',inplace=True)
data.loc[data.label==-1,'label']=1
#62天
print(np.unique(data.date))
num_all=data.groupby('date').apply(lambda x:len(x))
idx=np.arange(len(num_all))
p1=plt.bar(idx,num_all.values.reshape(-1))
plt.xlabel('date')
plt.ylabel('number')
plt.savefig('trade.jpg')
plt.clf()
num_bad=data[data.label==1].groupby('date').apply(lambda x:len(x))
idx=np.arange(len(num_bad))
p1=plt.bar(idx,num_bad.values.reshape(-1))
plt.xlabel('date')
plt.ylabel('number')
plt.savefig('bad.jpg')
plt.clf()
test=pd.read_csv('data/atec_anti_fraud_test_b.csv',index_col=0)
#f36-f47
#16627个bad f36-f47为nan
(data.f36==-1)&(data.f37==-1)&(data.f38==-1)&(data.f39==-1)&(data.f40==-1)&(data.f41==-1)&(data.f42==-1)&(data.f43==-1)&(data.f44==-1)&(data.f45==-1)&(data.f46==-1)&(data.f47==-1)
def select_sample():
test=pd.read_csv('data/atec_anti_fraud_test_b.csv',index_col=0)
train=pd.read_csv('data/atec_anti_fraud_train.csv',index_col=0)
train.fillna(-1,inplace=True)
#f1-f19 int 型
max_lst=[]
min_lst=[]
for i in range(20,298):
max_lst.append(np.max(test['f{}'.format(i)]))
min_lst.append(np.min(test['f{}'.format(i)]))
for index,(max_num,min_num) in enumerate(zip(max_lst,min_lst)):
train=train[((train['f{}'.format(index+20)]<=max_num) & (train['f{}'.format(index+20)]>=min_num)) | (train['f{}'.format(index+20)]==-1)]
return train
def vali(x,y):
lgbc=lgb.LGBMClassifier(n_estimators=500,max_depth=-1,num_leaves=100,learning_rate=0.01,subsample=0.8,sub_feature=0.8,random_state=0,n_jobs=-1,objective='binary',is_unbalance=True)
lgbc.fit(x,y)
precision_score(y,lgbc.predict(x))
roc_auc_score(y,lgbc.predict_proba(x)[:,1])
recall_score(y,lgbc.predict(x))
train_0=pd.read_csv('train_0.csv',index_col=0)
train_1=pd.read_csv('train_1.csv',index_col=0)
#train=train_0.iloc[np.random.choice(range(len(train_0)),size=13*len(train_1),replace=False)]
data=train_0.append(train_1)
X=data[data.columns.difference(['label'])]
X.drop('date',axis=1,inplace=True)
Y=data.label
lgbc=lgb.LGBMClassifier(n_estimators=500,max_depth=-1,class_weight={0:1,1:1.5},num_leaves=100,learning_rate=0.01,subsample=0.8,sub_feature=0.8,random_state=0,n_jobs=-1,objective='binary')
lgbc.fit(X,Y)
test=pd.read_csv('data/atec_anti_fraud_test_b.csv',index_col=0)
test.fillna(-1,inplace=True)
test.drop('date',axis=1,inplace=True)
pred_prob=lgbc.predict_proba(test)
temp=pd.DataFrame(pred_prob[:,1],index=test.index)
temp.reset_index(inplace=True)
temp.columns=['id','score']
temp.to_csv('test_wtf_b.csv',index=None)
if __name__=='__main__':
sub_pred()
#pass