-
Notifications
You must be signed in to change notification settings - Fork 34
/
001_model_lgb.py
146 lines (112 loc) · 5.48 KB
/
001_model_lgb.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
#coding:utf-8
import pandas as pd
import numpy as np
import time
import datetime
import gc
from sklearn.metrics import roc_auc_score, log_loss
from sklearn.preprocessing import LabelEncoder,OneHotEncoder
from scipy import sparse
import lightgbm as lgb
from sklearn.model_selection import GridSearchCV
import warnings
def ignore_warn(*args ,**kwargs):
pass
warnings.warn = ignore_warn
def logloss(act, pred):
epsilon = 1e-15
pred = sp.maximum(epsilon, pred)
pred = sp.minimum(1-epsilon, pred)
ll = sum(act*sp.log(pred) + sp.subtract(1,act)*sp.log(sp.subtract(1,pred)))
ll = ll * -1.0/len(act)
return ll
def lgb_model():
print('LGBMClassifier...')
lgb_clf = lgb.LGBMClassifier(boosting_type='gbdt', num_leaves=48, max_depth=-1, learning_rate=0.05, n_estimators=2000,
max_bin=425, subsample_for_bin=50000, objective='binary', min_split_gain=0,
min_child_weight=5, min_child_samples=10, subsample=1, subsample_freq=1,
colsample_bytree=1, reg_alpha=3, reg_lambda=5, seed=1000, nthread=-1, silent=True)
return lgb_clf
def feat_select(train, test):
features = train.drop(['is_trade', 'time', 'context_timestamp', 'day'], axis=1).columns.tolist()
target = ['is_trade']
return features, target
def lgbCV(train, test, All):
features, target = feat_select(train, test)
if All == False:
train_x = train.loc[train.hour < 10]
print(train_x.shape)
test_x = train.loc[train.hour >= 10]
print(test_x.shape)
if All == True:
train_x = train.loc[train.day < 7]
train_x1 = train.loc[(train.day == 7) & (train.hour < 10)]
train_x = pd.concat([train_x, train_x1])
print(train_x.shape)
test_x = train.loc[(train.day == 7) & (train.hour >= 10)]
print(test_x.shape)
del train_x1
gc.collect()
lgb_clf = lgb.LGBMClassifier(boosting_type='gbdt', num_leaves=48, max_depth=-1, learning_rate=0.05, n_estimators=2000,
max_bin=425, subsample_for_bin=50000, objective='binary', min_split_gain=0,
min_child_weight=5, min_child_samples=10, subsample=1, subsample_freq=1,
colsample_bytree=1, reg_alpha=3, reg_lambda=5, seed=1000, nthread=-1, silent=True)
lgb_model = lgb_clf.fit(train_x[features], train_x[target], eval_set=[(test_x[features], test_x[target])], early_stopping_rounds=200)
best_iter = lgb_model.best_iteration_
# 特征重要性
lgb_predictors = [i for i in train_x[features].columns]
lgb_feat_imp = pd.Series(lgb_model.feature_importances_, lgb_predictors).sort_values(ascending=False)
lgb_feat_imp.to_csv('lgb_feat_imp.csv')
# 训练模型
lgb_clf.fit(train_x[features], train_x[target])
test_x['lgb_predict'] = lgb_clf.predict_proba(test_x[features])[:, 1]
lgb_loss = log_loss(test_x[target], test_x['lgb_predict'])
lgb_auc = roc_auc_score(test_x[target], test_x['lgb_predict'])
print('Training loss: %.6f, Training AUC: %.6f' % (lgb_loss, lgb_auc))
return best_iter,lgb_loss
def sub(train, test, best_iter, lgb_loss):
features, target = feat_select(train, test)
lgb_clf = lgb.LGBMClassifier(boosting_type='gbdt', num_leaves=48, max_depth=-1, learning_rate=0.05, n_estimators=best_iter,
max_bin=425, subsample_for_bin=50000, objective='binary', min_split_gain=0,
min_child_weight=5, min_child_samples=10, subsample=1, subsample_freq=1,
colsample_bytree=1, reg_alpha=3, reg_lambda=5, seed=1000, nthread=-1, silent=True)
lgb_clf = lgb_model()
lgb_clf.fit(train[features], train[target])
test['predicted_score'] = lgb_clf.predict_proba(test[features])[:, 1]
now = datetime.datetime.now()
now = now.strftime('%m-%d-%H-%M')
mean = test['predicted_score'].mean()
print('predicted_score mean:', mean)
test[['instance_id', 'predicted_score']].to_csv("./submit/lgb_%s_%s.csv" % (round(lgb_loss,4)*10000, round(mean,5)*100000), index=False, sep=" ")
def main():
path = './data/'
All = False
if All == False:
data = pd.read_csv(path+"final_data_05-11-05-51.csv")
train = data.loc[data.hour < 12]
test = pd.read_csv(path+"test_day7.csv")
test = test[['instance_id']]
test = pd.merge(test, data, how='left', on='instance_id')
if All == True:
data = pd.read_csv(path+"all_final_data_05-11-06-46.csv")
train = pd.read_csv(path+"train_all.csv")
train = train[['instance_id']]
train =pd.merge(train, data, how='left', on='instance_id')
test = pd.read_csv(path+"test_all.csv")
test = test[['instance_id']]
test = pd.merge(test, data, how='left', on='instance_id')
print('The end of load data!')
print(train.shape)
print(test.shape)
del data
gc.collect()
print('Start predicting...')
print("----------------------------------------------------线下----------------------------------------")
best_iter, lgb_loss = lgbCV(train, test, All)
print("----------------------------------------------------线上----------------------------------------")
test_b = pd.read_table(path+'round2_ijcai_18_test_b_20180510.txt', sep=" ")
test_b = test_b[['instance_id']]
test = pd.merge(test_b, test, how='left', on='instance_id')
sub(train, test, best_iter, lgb_loss)
if __name__ == '__main__':
main()