-
Notifications
You must be signed in to change notification settings - Fork 0
/
inference.py
141 lines (118 loc) · 5.47 KB
/
inference.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Nov 28 11:00:00 2019
@author: tunan
"""
import os,json,pickle
import numpy as np
import pandas as pd
from lightgbm import LGBMClassifier
from xgboost import XGBClassifier
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import RobustScaler
from sklearn.linear_model import LogisticRegression
from sklearn.base import BaseEstimator, TransformerMixin, RegressorMixin, clone
from sklearn.ensemble import GradientBoostingClassifier,ExtraTreesClassifier,AdaBoostClassifier,RandomForestClassifier
from util import print_analyse,plot_four_roc
from dataset import dataset
from ml_model import StackingAveragedModels
class fit_ml_model(object):
#这个类只拟合模型，不寻找最优参数。
def __init__(self,base_model,name):
self.feature = None
self.label = None
self.model = None
self.base_model = base_model #未拟合前的模型。
self.log_path = "../log/{}_config.json".format(name) #最优参数的存储地址。
self.model_path = "../reject_models/{}_model".format(name) #模型地址修改，reject_models
self.name = name
if not os.path.exists("../reject_models"):
os.mkdir("../reject_models")
def load_model_with_params(self):
#读取文件中的最优参数，并返回设置好参数的模型
if not os.path.exists(self.log_path):
raise ValueError("no file in {}".format(self.log_path))
return None
self.result = json.load(open(self.log_path,'r'))
model = self.base_model(**self.result['params'])
return model
def fit(self,X_train,y_train):
'''
description:
1. 从文件中得到最优参数.
2. 最优参数的模型拟合特征和标签训练最优模型，并存入文件。
3. 清空特征和标签。
4. 如果已有模型文件就直接读出。
'''
if not os.path.exists(self.model_path):
self.feature = X_train
self.label = y_train
print('start training model!')
self.load_model_with_params()
model = self.base_model(**self.result['params'])
model.fit(X_train,y_train)
pickle.dump(model,open(self.model_path,'wb'))
self.feature = None
self.label = None
else:
print("loading model from file")
model = pickle.load(open(self.model_path,'rb'))
self.model = model
def predict_proba(self,X_test):
return self.model.predict_proba(X_test)
def _print_analyse(self,x_test,y_test,save_img = False):
y_pred = self.predict_proba(x_test)
print_analyse(y_test,y_pred[:,1],name = "reject_"+self.name)
if __name__ == "__main__":
reject_predict = pickle.load(open("../intermediate/reject_predict",'rb'))
data = dataset()
accept_sample,validation,reject_sample = data.run()
#缺失值填充
accept_sample.fillna(0,inplace = True)
validation.fillna(0,inplace = True)
reject_sample.fillna(0,inplace = True)
X_train = np.concatenate((accept_sample.values[:,2:],reject_sample.values[:,2:]),axis = 0)
y_train = np.concatenate((accept_sample.values[:,0],reject_predict),axis = 0)
X_test = validation.values[:,2:]
y_test = validation.values[:,0]
print("X_train shape is {}".format(X_train.shape))
print("y_train shape is {}".format(y_train.shape))
print("X_test shape is {}".format(X_test.shape))
print("y_test shape is {}".format(y_test.shape))
############# 模型拟合
#lightgbm
light_model = fit_ml_model(LGBMClassifier,name = 'lightgbm')
light_model.fit(X_train,y_train)
light_model._print_analyse(X_test,y_test,save_img = True)
#adaboost
ada_model = fit_ml_model(AdaBoostClassifier,name = 'adaboost')
ada_model.fit(X_train,y_train)
ada_model._print_analyse(X_test,y_test)
#gbdt
gbdt_model = fit_ml_model(GradientBoostingClassifier,name = 'gbdt')
gbdt_model.fit(X_train,y_train)
gbdt_model._print_analyse(X_test,y_test)
#random forest
rf_model = fit_ml_model(RandomForestClassifier,name = 'rf')
rf_model.fit(X_train,y_train)
rf_model._print_analyse(X_test,y_test)
#xgboost
xgb_model = fit_ml_model(XGBClassifier,name = 'xgboost')
xgb_model.fit(X_train,y_train)
xgb_model._print_analyse(X_test,y_test)
#stacking 提升法
base_model = [gbdt_model.load_model_with_params(),rf_model.load_model_with_params(),xgb_model.load_model_with_params()]
meta_model = make_pipeline(RobustScaler(),LogisticRegression(random_state=1,penalty = 'l2',solver = 'saga',max_iter = 100))
stack_model = StackingAveragedModels(base_models=base_model,meta_model=meta_model,name = 'reject_stacking')
stack_model.fit(X_train,y_train)
stack_model._print_analyse(X_test,y_test)
#四种模型的图像画在一起
rf_probas = rf_model.predict_proba(X_test)[:,1]
xgb_probas = xgb_model.predict_proba(X_test)[:,1]
gbdt_probas = gbdt_model.predict_proba(X_test)[:,1]
ada_probas = ada_model.predict_proba(X_test)[:,1]
stack_probas = stack_model.predict_proba(X_test)[:,1]
plot_four_roc(y_test,[rf_probas,xgb_probas,ada_probas,stack_probas],\
model_names = ['random forest','xgboost','gbdt','stacking model'],\
name = "ROC curve based on the rejected")