-
Notifications
You must be signed in to change notification settings - Fork 7
/
1_3_5_ROC.R
115 lines (113 loc) · 4.65 KB
/
1_3_5_ROC.R
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
setwd("E:/quanyuan_my/NG/53")
data_clean<-read.table("output7_0504_1.txt",sep = "\t")
#raw_data_all<-read.table("4-paras-53.txt",sep = "\t")
colnames(data_clean)<-c("x1","x2","x3","x4","lable")
#raw_data<-subset(raw_data_all,select = c(V3,V4,V7,V8,V9))
#colnames(raw_data)<-c("x1","x2","x3","x4","lable")
data_clean$lable<-factor(data_clean$lable)
select_2_func<-function(x,y,z){
p<-c()
for(i in 1:length(x)){
tmp<-c(x[i],y[i],z[i])
if(max(tmp)<0.5 | min(tmp)>0.5){
p[i]=mean(tmp)
}else{
if(length(tmp[which(tmp<0.5)])==1){
p[i]=mean(tmp[which(tmp>0.5)])
}else{
p[i]=mean(tmp[which(tmp<0.5)])
}
}
}
return(p)
}
#引入数据包
library(e1071)
library(caret)
library(pROC)
library(klaR)
library(ROCR)
library(gplots)
#1：1将数据分类
data_pos<-data_clean[data_clean$lable==1,]
data_nav<-data_clean[data_clean$lable==0,]
index<-sample(nrow(data_nav),nrow(data_pos))
data_clean_1<-rbind(data_nav[index,],data_pos)
data_clean_1<-data_clean_1[order(runif(nrow(data_clean_1))),]
auc1<-c()
for(i in 1:100){
folds<-createFolds(data_clean_1$lable,k=5)
prob_3models<-lapply(folds, function(x){
train_1<-data_clean_1[-x,]
test_1<-data_clean_1[x,]
svm1<-svm(lable~.,data = train_1,probability=T,gamma=1,cost=25)
glm1<-glm(lable~.,data = train_1,family=binomial)
nb1<-naiveBayes(lable~.,data = train_1)
glm_pred1<-predict(glm1,test_1,type="response")
nb_pred1<-c(predict(nb1,test_1,type = "raw")[,2])
svm_pred1<-predict(svm1,test_1,probability = T)
svm_pred1<-c(attr(svm_pred1,"probabilities")[,2])
return(select_2_func(glm_pred1,nb_pred1,svm_pred1))
})
prob_3models_all<-c(prob_3models$Fold1,prob_3models$Fold2,prob_3models$Fold3,prob_3models$Fold4,prob_3models$Fold5)
data_clean_1_fold<-rbind(data_clean_1[folds$Fold1,],data_clean_1[folds$Fold2,],data_clean_1[folds$Fold3,],data_clean_1[folds$Fold4,],data_clean_1[folds$Fold5,])
preed1=prediction(prob_3models_all,data_clean_1_fold$lable)
#pred.auc1<-performance(preed1,"tpr","fpr")
pred.auc1<-performance(preed1,"auc")
auc1[i]<-unlist(pred.auc1@y.values)
#plot(pred.auc1,type="l",lty=1)
}
#1：3将数据分类
index3<-sample(nrow(data_nav),nrow(data_pos)*3)
data_clean_3<-rbind(data_nav[index3,],data_pos)
data_clean_3<-data_clean_3[order(runif(nrow(data_clean_3))),]
auc3<-c()
for(i in 1:10){
folds3<-createFolds(data_clean_3$lable,k=5)
prob_3models_3<-lapply(folds3, function(x){
train_3<-data_clean_3[-x,]
test_3<-data_clean_3[x,]
svm3<-svm(lable~.,data = train_3,probability=T,gamma=1,cost=25)
glm3<-glm(lable~.,data = train_3,family=binomial)
nb3<-naiveBayes(lable~.,data = train_3)
glm_pred3<-predict(glm3,test_3,type="response")
nb_pred3<-c(predict(nb3,test_3,type = "raw")[,2])
svm_pred3<-predict(svm3,test_3,probability = T)
svm_pred3<-c(attr(svm_pred3,"probabilities")[,2])
return(select_2_func(glm_pred3,nb_pred3,svm_pred3))
})
prob_3models_all3<-c(prob_3models_3$Fold1,prob_3models_3$Fold2,prob_3models_3$Fold3,prob_3models_3$Fold4,prob_3models_3$Fold5)
preed3=prediction(prob_3models_all3,rbind(data_clean_3[folds3$Fold1,],data_clean_3[folds3$Fold2,],data_clean_3[folds3$Fold3,],data_clean_3[folds3$Fold4,],data_clean_3[folds3$Fold5,])$lable)
#pred.auc3<-performance(preed3,"tpr","fpr")
pred.auc3<-performance(preed3,"auc")
auc3[i]<-unlist(pred.auc3@y.values)
#plot(pred.auc3,add=T,lty=3)
}
index5<-sample(nrow(data_nav),nrow(data_pos)*5)
data_clean_5<-rbind(data_nav[index5,],data_pos)
data_clean_5<-data_clean_5[order(runif(nrow(data_clean_5))),]
auc5<-c()
for(i in 1:50){
folds5<-createFolds(data_clean_5$lable,k=5)
prob_3models_5<-lapply(folds5, function(x){
train_5<-data_clean_5[-x,]
test_5<-data_clean_5[x,]
svm5<-svm(lable~.,data = train_5,probability=T,gamma=1,cost=25)
glm5<-glm(lable~.,data = train_5,family=binomial)
nb5<-naiveBayes(lable~.,data = train_5)
glm_pred5<-predict(glm5,test_5,type="response")
nb_pred5<-c(predict(nb5,test_5,type = "raw")[,2])
svm_pred5<-predict(svm5,test_5,probability = T)
svm_pred5<-c(attr(svm_pred5,"probabilities")[,2])
return(select_2_func(glm_pred5,nb_pred5,svm_pred5))
})
prob_3models_all5<-c(prob_3models_5$Fold1,prob_3models_5$Fold2,prob_3models_5$Fold3,prob_3models_5$Fold4,prob_3models_5$Fold5)
preed5=prediction(prob_3models_all5,rbind(data_clean_5[folds5$Fold1,],data_clean_5[folds5$Fold2,],data_clean_5[folds5$Fold3,],data_clean_5[folds5$Fold4,],data_clean_5[folds5$Fold5,])$lable)
#pred.auc5<-performance(preed5,"tpr","fpr")
pred.auc5<-performance(preed5,"auc")
auc5[i]<-unlist(pred.auc5@y.values)
#plot(pred.auc1,type="l",lty=1,lwd=6)
#abline(a=0,b=1,lwd=2,lty=2)
#plot(pred.auc3,lty=1,lwd=6,type="l")
#plot(pred.auc5,lty=1,lwd=6,type="l")
}