-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathbaseline.py
252 lines (219 loc) · 11.1 KB
/
baseline.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
import pandas as pd
import numpy as np
import warnings
import math
import os
import lightgbm as lgb
from sklearn.model_selection import KFold,StratifiedKFold
from sklearn.metrics import roc_auc_score,f1_score,cohen_kappa_score
from sklearn.preprocessing import LabelEncoder
warnings.filterwarnings('ignore')
def gen_pesudo_data():
tmp_aum = pd.read_csv('./train/aum/aum_m10.csv')
tmp_aum[['X1', 'X2', 'X3', 'X4', 'X5', 'X6', 'X7', 'X8']] = np.nan
tmp_aum.to_csv('./train/aum/aum_m6.csv',index=None)
tmp_cunkuan = pd.read_csv('./train/cunkuan/cunkuan_m10.csv')
tmp_cunkuan[['C1', 'C2']] = np.nan
tmp_cunkuan.to_csv('./train/cunkuan/cunkuan_m6.csv',index=None)
tmp_beh = pd.read_csv('./train/behavior/behavior_m9.csv')
tmp_beh[['B1', 'B2','B3', 'B4','B5', 'B6','B7']] = np.nan
tmp_beh.to_csv('./train/behavior/behavior_m6.csv',index=None)
class GetData(object):
def load_aum_data(self,path='./',m=1,iftrain=True):
data = pd.read_csv(path + 'aum_m{}.csv'.format(m) )
m = m if iftrain else m+12
data.columns = ['cust_no', 'X1_{}'.format(m), 'X2_{}'.format(m), 'X3_{}'.format(m), 'X4_{}'.format(m),
'X5_{}'.format(m), 'X6_{}'.format(m), 'X7_{}'.format(m), 'X8_{}'.format(m)]
return data
def load_cunkuan_data(self,path='./',m=1,iftrain=True):
data = pd.read_csv(path + 'cunkuan_m{}.csv'.format(m) )
data['C3'] = data['C1'] / (1+data['C2'])
m = m if iftrain else m+12
data.columns = ['cust_no', 'C1_{}'.format(m), 'C2_{}'.format(m), 'C3_{}'.format(m)]
return data
def load_beh_data(self,path='./',m=1,iftrain=True):
data = pd.read_csv(path + 'behavior_m{}.csv'.format(m) )
m = m if iftrain else m+12
if m in [3,6,9,12,15]:
data['x'] = pd.to_datetime(data['B6']) + pd.tseries.offsets.QuarterEnd()
data['B6'] = (pd.to_datetime(data['x']) - pd.to_datetime(data['B6'])).dt.days # 该季度与最后一次交易时间距离的天数
data = data.drop('x',axis=1)
data.columns = ['cust_no', 'B1_{}'.format(m), 'B2_{}'.format(m), 'B3_{}'.format(m), 'B4_{}'.format(m),
'B5_{}'.format(m), 'B6_{}'.format(m),'B7_{}'.format(m)]
else:
data['B6'] = np.nan
data['B7'] = np.nan
data.columns = ['cust_no', 'B1_{}'.format(m), 'B2_{}'.format(m), 'B3_{}'.format(m), 'B4_{}'.format(m),
'B5_{}'.format(m),'B6_{}'.format(m),'B7_{}'.format(m)]
return data
def get_data(data_file='',param=''):
df_aum = pd.DataFrame()
for mon in range(3,0,-1):
df = getattr(GetData(), param)(path='./test/{}/'.format(data_file),m= mon,iftrain=False)
if len(df_aum):
df_aum = df_aum.merge(df,on='cust_no',how='left')
else:
df_aum = df
for mon in range(12,5,-1):
df = getattr(GetData(), param)(path='./train/{}/'.format(data_file),m= mon,iftrain=True)
df_aum = df_aum.merge(df,on='cust_no',how='left')
return df_aum
def get_quater_data(df,data_type = '',colhead='X',num_col=3,quater=3):
if data_type == 'aum':
columns = ['cust_no']
rename_col = ['cust_no']
for i in range(1,num_col + 1):
for j in range(4):
columns.append('{}{}_{}'.format(colhead,i,quater*3-j))
rename_col.append('{}{}_{}'.format(colhead,i,4 - j))
tmp = df[columns]
tmp.columns = rename_col
return tmp
def statistics_feature_aum(df):
for i in range(1,9):
df['X{}_mean'.format(i)] = df[['X{}_4'.format(i),'X{}_3'.format(i),'X{}_2'.format(i)]].mean(axis=1)
df['X{}_std'.format(i)] = df[['X{}_4'.format(i),'X{}_3'.format(i),'X{}_2'.format(i)]].std(axis=1)
df['X{}_sum'.format(i)] = df[['X{}_4'.format(i),'X{}_3'.format(i),'X{}_2'.format(i)]].sum(axis=1)
df['X{}_max'.format(i)] = df[['X{}_4'.format(i),'X{}_3'.format(i),'X{}_2'.format(i)]].max(axis=1)
df['X{}_min'.format(i)] = df[['X{}_4'.format(i),'X{}_3'.format(i),'X{}_2'.format(i)]].min(axis=1)
return df
def aum_feat_engineering(df):
df['X3_sub1'] = df['X3_4'] - df['X3_3']
df['X3_sub2'] = df['X3_4'] - df['X3_2']
df['X3_sub3'] = df['X3_3'] - df['X3_2']
df['X3_sub4'] = df['X3_2'] - df['X3_1']
df['X3_sub5'] = df['X3_3'] - df['X3_1']
df['X1_sub1'] = df['X1_4'] - df['X1_3']
df['X1_sub2'] = df['X1_4'] - df['X1_2']
df['X1_sub3'] = df['X1_4'] - df['X1_1']
df['X1_sub4'] = df['X1_4'] - df['X1_1']
df['X7_sub1'] = df['X7_4'] - df['X7_3']
df['X7_sub2'] = df['X7_4'] - df['X7_2']
df['X7_sub3'] = df['X7_4'] - df['X7_1']
df['sum_X123_4'] = df[['X1_4','X2_4','X3_4']].sum(axis=1)
df['sum_X123_3'] = df[['X1_3','X2_3','X3_3']].sum(axis=1)
df['sum_X123_2'] = df[['X1_2','X2_2','X3_2']].sum(axis=1)
df['sum_X123_1'] = df[['X1_1','X2_1','X3_1']].sum(axis=1)
df['X123_sub1'] = df['sum_X123_4'] - df['sum_X123_3']
df['X123_sub2'] = df['sum_X123_4'] - df['sum_X123_2']
df['X123_sub3'] = df['sum_X123_4'] - df['sum_X123_1']
df['X7_div_X123_sum'] = df['X7_4'] / (1e-3 + df['sum_X123_4'])
df['X7_sub_X123_sum'] = df['X7_4'] - df['sum_X123_4']
return df
def eval_error(pred,train_set):
labels = train_set.get_label()
pred = pred.reshape((3,int(len(pred)/3))).T
y_pred = pred.argmax(axis=1)
score = cohen_kappa_score(labels,y_pred)
return 'kappa_score',score,True
def model_train(df,trainlabel,cate_cols,test_,feature,num_class):
'''
@param df: 训练数据 DataFrame
@param trainlabel：训练标签 string eg. 'label'
@param cate_cols: 类别变量名 list eg. ['col1','col2'...]
@param test_ : 测试数据 DataFrame
@param feature ：所有训练特征 list eg. ['feat1','feat2'...]
@return sub_preds: 预测数据
'''
train_= df.copy()
auc = []
n_splits = 5
oof_lgb = np.zeros([len(train_),num_class])
folds = KFold(n_splits=n_splits, shuffle=True, random_state=2019)
stratifiedKfold = StratifiedKFold(n_splits=n_splits, shuffle=True, random_state=2019)
sub_preds = np.zeros([test_.shape[0],num_class])
sub_preds1 = np.zeros([test_.shape[0],n_splits])
use_cart = True
cate_cols = cate_cols
label = trainlabel
pred = list(feature)
params = {
'learning_rate': 0.02,
'boosting_type': 'gbdt',
'objective':'multiclass',
#'metric':'multi-error',
'num_class':num_class,
'num_leaves':60,
'feature_fraction': 0.8,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'seed': 1,
'bagging_seed': 1,
'feature_fraction_seed': 5,
'min_data_in_leaf': 20,
'max_depth':-1,
'nthread': 8,
'verbose': 1,
# 'is_unbalanace':True,
# 'lambda_l1': 0.4,
# 'lambda_l2': 0.5,
'device': 'gpu'
}
for n_fold, (train_idx, valid_idx) in enumerate(folds.split(train_[pred],train_[[label]]), start=1):
print('the %s training start ...'%n_fold)
train_x, train_y,train_weight = train_[pred].iloc[train_idx], train_[[label]].iloc[train_idx],train_[['weight']].iloc[train_idx]
valid_x, valid_y,valid_weight = train_[pred].iloc[valid_idx], train_[[label]].iloc[valid_idx],train_[['weight']].iloc[valid_idx]
#x = train_['bid'].iloc[valid_idx]
print(train_y.shape)
if use_cart:
dtrain = lgb.Dataset(train_x, label=train_y, categorical_feature=cate_cols,weight=train_weight.values.flatten(order='F'))
dvalid = lgb.Dataset(valid_x, label=valid_y, categorical_feature=cate_cols)
#dvalid1 = lgb.Dataset(test_[pred], label=test_[['label']], categorical_feature=cate_cols)
else:
dtrain = lgb.Dataset(train_x, label= train_y)
dvalid = lgb.Dataset(valid_x, label= valid_y)
clf = lgb.train(
params=params,
train_set=dtrain,
num_boost_round=1000,
valid_sets=[dvalid],
# early_stopping_rounds = 100,
verbose_eval=100
,feval=eval_error
)
sub_preds += clf.predict(test_[pred].values,num_iteration=1000)/ folds.n_splits
sub_preds1[:,n_fold-1] = clf.predict(test_[pred].values,num_iteration=400).argmax(axis=1)
train_pred = clf.predict(valid_x,num_iteration=clf.best_iteration)
y_pred = train_pred.argmax(axis=1)
oof_lgb[valid_idx] = train_pred
#print('MEAN AUC:',np.mean(auc))
return sub_preds,oof_lgb,clf,sub_preds1
def main():
gen_pesudo_data()
df_aum = get_data(data_file='aum',param='load_aum_data')
df_aum_test = get_quater_data(df_aum,data_type='aum',quater=5,colhead='X',num_col=8)
df_aum_test_Q4 = get_quater_data(df_aum,data_type='aum',quater=4,colhead='X',num_col=8)
df_aum_test_Q3 = get_quater_data(df_aum,data_type='aum',quater=3,colhead='X',num_col=8)
df_aum_test = aum_feat_engineering(df_aum_test)
df_aum_test_Q4 = aum_feat_engineering(df_aum_test_Q4)
df_aum_test_Q3 = aum_feat_engineering(df_aum_test_Q3)
df_aum_test = statistics_feature_aum(df_aum_test)
df_aum_test_Q4 = statistics_feature_aum(df_aum_test_Q4)
df_aum_test_Q3 = statistics_feature_aum(df_aum_test_Q3)
cust_avli_Q3 = pd.read_csv('./train/avli/cust_avli_Q3.csv')
cust_avli_Q4 = pd.read_csv('./train/avli/cust_avli_Q4.csv')
cust_avli_test = pd.read_csv('./test/avli/cust_avli_Q1.csv')
cust_avli_Q3 = df_aum_test_Q3.loc[df_aum_test_Q3.cust_no.isin(cust_avli_Q3.cust_no.values)]
cust_avli_Q4 = df_aum_test_Q4.loc[df_aum_test_Q4.cust_no.isin(cust_avli_Q4.cust_no.values)]
cust_avli_test = df_aum_test.loc[df_aum_test.cust_no.isin(cust_avli_test.cust_no.values)]
label_Q3 = pd.read_csv('./train/y/y_Q3_3.csv')
label_Q4 = pd.read_csv('./train/y/y_Q4_3.csv')
label_Q3['label'] = label_Q3.label + 1
label_Q4['label'] = label_Q4.label + 1
cust_avli_Q3 = cust_avli_Q3.merge(label_Q3,on='cust_no',how='left')
cust_avli_Q4 = cust_avli_Q4.merge(label_Q4,on='cust_no',how='left')
cust_avli_Q3['pre_label'] = np.nan
last_label = label_Q3[['cust_no','label']]
last_label.columns = ['cust_no','pre_label']
cust_avli_Q4 = cust_avli_Q4.merge(last_label,on='cust_no',how='left')
last_label = label_Q4[['cust_no','label']]
last_label.columns = ['cust_no','pre_label']
cust_avli_test = cust_avli_test.merge(last_label,on='cust_no',how='left')
Train_data = pd.concat([cust_avli_Q3,cust_avli_Q4],axis=0)
Train_data['weight'] = Train_data.label.map({1:1.03,2:0.58,0:1})
feature = Train_data.drop(['cust_no','label','weight'],axis=1).columns
sub_preds,oof_lgb,clf,sub_preds1 = model_train(Train_data,trainlabel='label',cate_cols=[],test_=cust_avli_test,feature=feature,num_class=3)
cust_avli_test['label'] = sub_preds.argmax(axis=1) - 1
cust_avli_test[['cust_no','label']].to_csv('./test/baseline.csv',index=None)
if __name__ == "__main__":
main()