-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathstacking_class.py
190 lines (185 loc) · 9.03 KB
/
stacking_class.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
#--coding=utf-8--
# coding=utf8
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt # for plotting
from scipy.stats import mode
import warnings # to ignore warnings
from sklearn import preprocessing
from sklearn.externals import joblib
from sklearn.metrics import mean_squared_error #均方误差
from sklearn.metrics import mean_absolute_error #平方绝对误差
from sklearn.metrics import r2_score#R square
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GridSearchCV
import re
import seaborn as sns
from xgboost import XGBRegressor
from sklearn import datasets
from sklearn.ensemble import RandomForestRegressor, ExtraTreesRegressor, GradientBoostingRegressor
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import StratifiedKFold
import numpy as np
from sklearn.metrics import roc_auc_score
from sklearn.datasets.samples_generator import make_blobs
from xgboost import XGBRegressor
from sklearn.linear_model import LogisticRegression
'''创建训练的数据集'''
from sklearn.model_selection import KFold
'''5折stacking'''
class stacking_model:
def __init__(self):
print(1)
def first_level_model(self,nfolds,train_data,target_data,test_data,clfs):#n_folds:K折次数，train_data:训练集,target_data:目标特征值列表，test_data:测试数据集，clfs：stacking基模型列表
n_folds = 5
train_set = []
test_set = []
target_set = []
new_target_value = []
new_feature_value = []
new_feature_all = []
new_train_value = []
test_data_set = []
test_id = []
skf = KFold(n_splits=nfolds, shuffle=False).split(train_data)
print(train_data[3].shape)
print(train_data[3])
for train, test in skf:
for j, clf in enumerate(clfs):
train_set = list(map(lambda a: list(train_data[a, :]), train))
test_set = list(map(lambda a: list(train_data[a, :]), test))
target_set = list(map(lambda a: target_data[a], train))
train_set = pd.DataFrame(train_set)
print(train_set.shape)
print(len(target_set))
print("model", clf)
clf.fit(train_set, target_set)
new_feature_target = clf.predict(test_set)
new_feature_value.append((j, new_feature_target))
test_id.append((j, test))
test_data_set.append((j, clf.predict(test_data)))
print("新的特征", test_id)
print("新的特征序列", new_feature_value)
print("测试数据集", test_data_set)
return new_feature_value,test_data_set
def second_model_train(self,new_feature_value,test_data_set):#返回基模型处理后的超特征矩阵，和测试数据集融合特征
new_feature_0 = []
new_feature_1 = []
new_feature_2 = []
new_feature_3 = []
new_feature_4 = []
new_feature_matrix = []
for i in new_feature_value:
if i[0] == 0:
new_feature_0 = new_feature_0 + list(i[1])
elif i[0] == 1:
new_feature_1 = new_feature_1 + list(i[1])
elif i[0] == 2:
new_feature_2 = new_feature_2 + list(i[1])
elif i[0] == 3:
new_feature_3 = new_feature_3 + list(i[1])
elif i[0] == 4:
new_feature_4 = new_feature_4 + list(i[1])
print(new_feature_1)
new_feature_matrix.append(new_feature_0)
new_feature_matrix.append(new_feature_1)
new_feature_matrix.append(new_feature_2)
new_feature_matrix.append(new_feature_3)
new_feature_matrix.append(new_feature_4)
new_feature_matrix = pd.DataFrame(new_feature_matrix)
new_feature_matrix = new_feature_matrix.T
print(new_feature_matrix)
new_test_0 = []
new_test_1 = []
new_test_2 = []
new_test_3 = []
new_test_4 = []
new_test_matrix = []
for i in test_data_set:
if i[0] == 0:
new_test_0.append(list(i[1]))
elif i[0] == 1:
new_test_1.append(list(i[1]))
elif i[0] == 2:
new_test_2.append(list(i[1]))
elif i[0] == 3:
new_test_3.append(list(i[1]))
elif i[0] == 4:
new_test_4.append(list(i[1]))
new_test_0 = (pd.DataFrame(new_test_0)).T
new_test_1 = (pd.DataFrame(new_test_1)).T
new_test_2 = (pd.DataFrame(new_test_2)).T
new_test_3 = (pd.DataFrame(new_test_3)).T
new_test_4 = (pd.DataFrame(new_test_4)).T
j = 0
B0 = []
B1 = []
B2 = []
B3 = []
B4 = []
new_test = [new_test_0, new_test_1, new_test_2, new_test_3, new_test_4]
for i in range(len(new_test)):
for j in range(len(new_test[i])):
if i == 0:
B0.append(np.mean(new_test[i].iloc[j, :]))
elif i == 1:
B1.append(np.mean(new_test[i].iloc[j, :]))
elif i == 2:
B2.append(np.mean(new_test[i].iloc[j, :]))
elif i == 3:
B3.append(np.mean(new_test[i].iloc[j, :]))
elif i == 4:
B4.append(np.mean(new_test[i].iloc[j, :]))
new_test_matrix.append(B0)
new_test_matrix.append(B1)
new_test_matrix.append(B2)
new_test_matrix.append(B3)
new_test_matrix.append(B4)
new_test_matrix = pd.DataFrame(new_test_matrix)
new_test_matrix = new_test_matrix.T
print(new_test_matrix)
return new_feature_matrix,new_test_matrix
def second_model_predict(self,new_feature_matrix,new_test_matrix,id):
other_params = {'learning_rate': 0.1, 'n_estimators': 200, 'max_depth': 3, 'min_child_weight': 5, 'seed': 0,
'subsample': 0.8, 'colsample_bytree': 0.8, 'gamma': 0.1, 'reg_alpha': 0, 'reg_lambda': 1,
'eval_metric': "mae", 'verbose': 1}
clf_2 = XGBRegressor(**other_params)
clf_2.fit(new_feature_matrix, target_data)
test_pre = clf_2.predict(new_test_matrix)
test_pre = [int(a) for a in test_pre]
print(len(test_pre))
dataframe = pd.DataFrame({'id': id, 'score': test_pre})
dataframe.to_csv("result_ensemble_stack.csv", index=False, sep=',')
if __name__ == '__main__':
warnings.filterwarnings('ignore')
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
data = pd.read_csv('train_dataset.csv', encoding='utf-8')
test = pd.read_csv('test_dataset.csv', encoding='utf-8')
data = data
data = data.fillna(0)
data_hot = data.drop(columns=["用户编码", "信用分", "是否大学生客户", "用户实名制是否通过核实", "当月是否到过福州山姆会员店", "用户最近一次缴费距今时长（月）",
"当月是否逛过福州仓山万达", "是否黑名单客户", "当月物流快递类应用使用次数",
"是否经常逛商场的人", "是否4G不健康客户", "当月火车类应用使用次数", "当月是否看电影", "当月是否体育场馆消费"])
data_hot = pd.DataFrame(data_hot, dtype=float)
train_columns = data_hot.columns.values
test_data = test.drop(columns=["用户编码", "是否大学生客户", "用户实名制是否通过核实", "当月是否到过福州山姆会员店", "用户最近一次缴费距今时长（月）",
"当月是否逛过福州仓山万达", "是否黑名单客户", "当月物流快递类应用使用次数",
"是否经常逛商场的人", "是否4G不健康客户", "当月火车类应用使用次数", "当月是否看电影", "当月是否体育场馆消费"])
test_data = pd.DataFrame(test_data, dtype=float)
target_data = data["信用分"]
train_data = preprocessing.scale(data_hot)
test_data = preprocessing.scale(test_data)
ran_params = {'criterion': 'mae', 'max_depth': 3, 'n_estimators': 200, 'verbose': 1}
logistic_params = {'penalty': 'l2', 'solver': 'lbfgs', 'verbose': 1}
'''模型融合中使用到的各个单模型'''
clfs = [LinearRegression(),
ExtraTreesRegressor(**ran_params, bootstrap=True),
RandomForestRegressor(**ran_params),
GradientBoostingRegressor(**ran_params),
LogisticRegression(**logistic_params)]
id = test["用户编码"]
new_feature_value, test_data_set=stacking_model().first_level_model(nfolds=5, train_data=train_data,target_data=target_data,test_data=test_data,clfs=clfs)
new_feature_matrix, new_test_matrix=stacking_model().second_model_train(new_feature_value=new_feature_value,test_data_set=test_data_set)
stacking_model().second_model_predict(new_feature_matrix=new_feature_matrix,new_test_matrix=new_test_matrix,id=id)