library(caret)
library(ROCR)
library(pROC)
library(e1071)
library(rpart)
set.seed(123)
Difteri<-read.table("D:/Machine Learning/dataDifteri.csv",sep=",",header=TRUE)
Y<-as.factor(Difteri$Y)
Difteri<-cbind(Y,Difteri[,-1])
part<-createDataPartition(Difteri$Y,p=0.80,list=FALSE)
training<-Difteri[part,]
testing<-Difteri[-part,]
reglog<-glm(Y~.,data=training,family=binomial(link="logit"))


predTr_reglog <- predict(reglog,newdata=training,type='response')
predTr_reglog<- ifelse(predTr_reglog>0.5,1,0)
conMat_Tr_reglog<-confusionMatrix(predTr_reglog,training$Y)
predTr2_reglog <- prediction(predTr_reglog, training$Y)
perfTr_reglog <- performance(predTr2_reglog, measure = "tpr", x.measure = "fpr")
plot(perfTr_reglog)


predTe_reglog <- predict(reglog,newdata=testing,type='response')
predTe_reglog<- ifelse(predTe_reglog>0.5,0,1)
conMat_Te_reglog<-confusionMatrix(predTe_reglog,testing$Y)
predTe2_reglog <- prediction(predTe_reglog, testing$Y)
perfTe_reglog <- performance(predTe2_reglog, measure = "tpr", x.measure = "fpr")
plot(perfTe_reglog)

##############SVM Radial#################
svmR<-svm(Y~.,data=training,type="C-classification",cross=10,probability=TRUE,kernel="radial")

predTr_svmR<-predict(svmR,training[,-1],probability=TRUE)
conMat_Tr_svmR<-confusionMatrix(predTr_svmR,training$Y)
predTr2_svmR<-prediction(attr(predTr_svmR,"probabilities")[,2], training[,1] == "0")
plot(performance(predTr2_svmR,"tpr","fpr"))
abline(0,1,lty=2,col="red")

predTe_svmR<-predict(svmR,testing[,-1],probability=TRUE)
conMat_Te_svmR<-confusionMatrix(predTe_svmR,testing$Y)
predTe2_svmR<-prediction(attr(predTe_svmR,"probabilities")[,2], testing[,1] == "0")
plot(performance(predTe2_svmR,"tpr","fpr"))
abline(0,1,lty=2,col="red")


#################CART#################
library(rpart)
cart<-rpart(Y~.,method="class",data=training,,control = rpart.control(minsplit = 20, cp = 1e-04))
plot(cart, uniform=TRUE, main="Classification Tree for Difteri")
text(cart, use.n=TRUE, all=TRUE, cex=.8)

predTr_cart<-predict(cart,training,type="class")
conMat_Tr_cart<-confusionMatrix(predTr_cart,training$Y)
predTr_cart<-predict(cart,training,type="prob")[,2]
predTr2_cart<-prediction(predTr_cart,training$Y)
plot(performance(predTr2_cart,"tpr","fpr"),main="ROC curve")
abline(0,1,lty=2,col="green")



predTe_cart<-predict(cart,testing,type="class")
conMat_Te_cart<-confusionMatrix(predTe_cart,testing$Y)
predTe_cart<-predict(cart,testing,type="prob")[,2]
predTe2_cart<-prediction(predTe_cart,testing$Y)
plot(performance(predTe2_cart,"tpr","fpr"),main="ROC curve")
abline(0,1,lty=2,col="green")


####################TRAINING DATA#########################
predTr_reglog <- predict(reglog,newdata=training,type='response')
predTr_reglog<- ifelse(predTr_reglog>0.5,0,1)
conMat_Tr_reglog<-confusionMatrix(predTr_reglog,training$Y)
predTr2_reglog <- prediction(predTr_reglog, training$Y)
perfTr_reglog <- performance(predTr2_reglog, measure = "tpr", x.measure = "fpr")


predTr_svmR<-predict(svmR,training[,-1],probability=TRUE)
conMat_Tr_svmR<-confusionMatrix(predTr_svmR,training$Y)
predTr2_svmR<-prediction(attr(predTr_svmR,"probabilities")[,2], training[,1] == "1")


predTr_cart<-predict(cart,training,type="class")
conMat_Tr_cart<-confusionMatrix(predTr_cart,training$Y)
predTr_cart<-predict(cart,training,type="prob")[,2]
predTr2_cart<-prediction(predTr_cart,training$Y)

plot(perfTr_reglog,lwd=2,main="ROC Training Data",lty=1,col="yellow")
abline(0,1,lty=2,col="black")
plot(performance(predTr2_svmR,"tpr","fpr"),lwd=2,add=TRUE,lty=1,col="green")
plot(performance(predTr2_cart,"tpr","fpr"),lwd=2,add=TRUE,lty=1,col="red")
rg<-range(0,0.5,0.5,0.5,0)
legend(0.5,rg[2],c("Reglog","SVM","CART"),title="Model",cex=0.8,col=c("yellow","green","red"),lwd=c(3,3,3),lty=c(1,1,1),box.lty=2,box.col="blue")

AUC1<-performance(predTr2_reglog,"auc")
AUC2<-performance(predTr2_svmR,"auc")
AUC3<-performance(predTr2_cart,"auc")


#######################TESTING DATA###################
predTe_reglog <- predict(reglog,newdata=testing,type='response')
predTe_reglog<- ifelse(predTe_reglog>0.5,0,1)
conMat_Te_reglog<-confusionMatrix(predTe_reglog,testing$Y)
predTe2_reglog <- prediction(predTe_reglog, testing$Y)
perfTe_reglog <- performance(predTe2_reglog, measure = "tpr", x.measure = "fpr")
plot(perfTe_reglog,main="ROC Testing Data",col="yellow")
abline(0,1,lty=2,col="black")

predTe_svmR<-predict(svmR,testing[,-1],probability=TRUE)
conMat_Te_svmR<-confusionMatrix(predTe_svmR,testing$Y)
predTe2_svmR<-prediction(attr(predTe_svmR,"probabilities")[,2], testing[,1] == "1")
plot(performance(predTe2_svmR,"fpr","tpr"),add=TRUE,col="green")


predTe_cart<-predict(cart,testing,type="class")
conMat_Te_cart<-confusionMatrix(predTe_cart,testing$Y)
predTe_cart<-predict(cart,testing,type="prob")[,2]
predTe2_cart<-prediction(predTe_cart,testing$Y)
plot(performance(predTe2_cart,"tpr","fpr"),add=TRUE,col="red")




plot(perfTe_reglog,lwd=2,main="ROC Testing Data",lty=1,col="yellow")
abline(0,1,lty=2,col="black")
plot(performance(predTe2_svmR,"tpr","fpr"),lwd=2,add=TRUE,lty=1,col="green")
plot(performance(predTe2_cart,"tpr","fpr"),lwd=2,add=TRUE,lty=1,col="red")
rg<-range(0,0.5,0.5,0.5,0)
legend(0.5,rg[2],c("Reglog","SVM","CART"),title="Model",cex=0.8,col=c("yellow","green","red"),lwd=c(3,3,3),lty=c(1,1,1),box.lty=2,box.col="blue")

AUC1<-performance(predTe2_reglog,"auc")
AUC2<-performance(predTe2_svmR,"auc")
AUC3<-performance(predTe2_cart,"auc")




Difteri<-read.table("D:/Machine Learning/dataDifteri.csv",sep=",",header=TRUE)
Y<-as.factor(Difteri$Y)
Difteri<-cbind(Y,Difteri[,-1])
part<-createDataPartition(Difteri$Y,p=0.80,list=FALSE)
training<-Difteri