-
Notifications
You must be signed in to change notification settings - Fork 84
/
data_pred_a+b.py
145 lines (124 loc) · 6.17 KB
/
data_pred_a+b.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
from tool import *
from feat_a import *
data_path = 'data/b/'
d = {89950166: 1, 89950167: 2, 89950168: 5, 90063345: 0, 90109916: 4,
90155946: 8, 99999825: 10, 99999826: 7, 99999827: 6, 99999828: 3, 99999830: 9}
rd = {0: 90063345, 1: 89950166, 2: 89950167, 3: 99999828, 4: 90109916,
5: 89950168, 6: 99999827, 7: 99999826, 8: 90155946, 9: 99999830, 10: 99999825}
d1 = {0: 0, 4: 1, 8: 2}
rd1 = {0: 0, 1: 4, 2: 8}
d4 = {1: 0, 2: 1, 3: 2, 5: 3, 6: 4, 7: 5, 9: 6, 10: 7}
rd4 = {0: 1, 1: 2, 2: 3, 3: 5, 4: 6, 5: 7, 6: 9, 7: 10}
cc = ['service_type', 'is_mix_service', 'is_promise_low_consume',
'net_service', 'gender', 'age', 'online_time', 'contract_type',
'1_total_fee', '2_total_fee', '3_total_fee', '4_total_fee',
'month_traffic', 'last_month_traffic', 'local_trafffic_month',
'local_caller_time', 'service1_caller_time', 'service2_caller_time',
'many_over_bill', 'contract_time', 'pay_times', 'pay_num']
#print('读取train数据...')
train_old = pd.read_csv('data/a/' + 'train.csv')
train = pd.read_csv(data_path + 'train_new.csv')
test = pd.read_csv(data_path + 'test_new.csv')
data = train_old.append(train).drop_duplicates(cc).append(test)
#print('构造特征...')
data_feat = make_feat(data, 'online')
test_feat = data_feat[data_feat['user_id'].isin(test['user_id'])].copy()
train_feat = data_feat[data_feat['user_id'].isin(train['user_id'])].copy()
train_feat1 = train_feat[train_feat['service_type'] == 1].copy()
test_feat1 = test_feat[test_feat['service_type'] == 1].copy()
train_feat1['label'] = train_feat1['label'].map(d1)
predictors1 = [c for c in train_feat.columns if (c not in ['user_id', 'current_service', 'label']) and
('contract_type' not in c) and ('service_type' not in c)]
params = {'objective': 'multi:softprob',
'eta': 0.02,
'max_depth': 6,
'silent': 1,
'num_class': 3,
'eval_metric': "mlogloss",
'min_child_weight': 3,
'subsample': 0.7,
'colsample_bytree': 0.7,
'seed': 66
}
train = pd.read_csv(data_path + 'train_new.csv')
def xgb_cv(params, train_feat, test_feat, predictors, label='label', cv=5, stratified=True):
#print('开始CV 5折训练...')
t0 = time.time()
train_preds = np.zeros((len(train_feat), train_feat[label].nunique()))
test_preds = np.zeros((len(test_feat), train_feat[label].nunique()))
xgb_test = xgb.DMatrix(test_feat[predictors])
models = []
if stratified:
folds = StratifiedKFold(n_splits=cv, shuffle=True, random_state=66)
else:
folds = KFold(n_splits=cv, shuffle=True, random_state=66)
for i, (train_index, test_index) in enumerate(folds.split(train_feat, train_feat[label])):
temp = train_feat.iloc[train_index]
# temp = temp.append(temp[temp['user_id'].isin(train['user_id'].values)])
xgb_train = xgb.DMatrix(temp[predictors], temp['label'])
temp = train_feat.iloc[test_index]
temp = temp[temp['user_id'].isin(train['user_id'].values)]
xgb_eval_temp = xgb.DMatrix(temp[predictors], temp['label'])
xgb_eval = xgb.DMatrix(train_feat[predictors].iloc[test_index], train_feat['label'].iloc[test_index])
#print('开始第{}轮训练...'.format(i))
params = {'objective': 'multi:softprob',
'eta': 0.02,
'max_depth': 6,
'silent': 1,
'num_class': 11,
'eval_metric': "mlogloss",
'min_child_weight': 3,
'subsample': 0.7,
'colsample_bytree': 0.7,
'seed': 66
} if params is None else params
watchlist = [(xgb_train, 'train'), (xgb_eval_temp, 'val')]
clf = xgb.train(params,
xgb_train,
num_boost_round=3000,
evals=watchlist,
verbose_eval=50,
early_stopping_rounds=50)
train_preds[test_index] += clf.predict(xgb_eval)
test_preds += clf.predict(xgb_test)
models.append(clf)
#print('用时{}秒'.format(time.time() - t0))
return train_preds, test_preds / 5
train_preds1, test_preds1 = xgb_cv(params, train_feat1, test_feat1, predictors1)
int_train_preds1 = train_preds1.argmax(axis=1)
int_test_preds1 = test_preds1.argmax(axis=1)
#print('线下第一类的得分为： {}'.format(multi_f1(train_feat1['label'], int_train_preds1) ** 2))
train_preds1 = pd.DataFrame(train_preds1)
train_preds1['user_id'] = train_feat1['user_id'].values
test_preds1 = pd.DataFrame(test_preds1)
test_preds1['user_id'] = test_feat1['user_id'].values
data_pred1 = train_preds1.append(test_preds1)
data_pred1.columns = [rd1[i] if i in rd1 else i for i in data_pred1.columns]
train_feat4 = train_feat[train_feat['service_type'] != 1].copy()
test_feat4 = test_feat[test_feat['service_type'] != 1].copy()
train_feat4['label'] = train_feat4['label'].map(d4)
predictors4 = [c for c in train_feat.columns if (c not in ['user_id', 'current_service', 'label'])]
params = {'objective': 'multi:softprob',
'eta': 0.02,
'max_depth': 6,
'silent': 1,
'num_class': 8,
'eval_metric': "mlogloss",
'min_child_weight': 3,
'subsample': 0.7,
'colsample_bytree': 0.7,
'seed': 66
}
train_preds4, test_preds4 = xgb_cv(params, train_feat4, test_feat4, predictors4)
int_train_preds4 = train_preds4.argmax(axis=1)
int_test_preds4 = test_preds4.argmax(axis=1)
#print('线下第四类的得分为： {}'.format(multi_f1(train_feat4['label'], int_train_preds4) ** 2))
train_preds4 = pd.DataFrame(train_preds4)
train_preds4['user_id'] = train_feat4['user_id'].values
test_preds4 = pd.DataFrame(test_preds4)
test_preds4['user_id'] = test_feat4['user_id'].values
data_pred4 = train_preds4.append(test_preds4)
data_pred4.columns = [rd4[i] if i in rd4 else i for i in data_pred4.columns]
# 输出预测概率，做stacking使用
data_pred = data_pred1.append(data_pred4).fillna(0)
data_pred.to_csv('data_preds_xgb1_20181030_050913.csv', index=False)