deeper.tree<- rpart(INCOME_C4 ~ ., data = train.data3[,-1], method = "class", cp = .00001, minsplit =5)

options(digits=8)

printcp(deeper.tree)

which.min(deeper.tree$cptable[,"xerror"])

options(digits=8)

X.E=deeper.tree$cptable[8:19,"xerror"]

options(digits=8)

X.S=deeper.tree$cptable[8:19,"xstd"]

CP.OPT=X.E+X.S

which.min(CP.OPT)

OPT.POINT=abs(deeper.tree$cptable[c(1:9,11:29),"xerror"] -deeper.tree$cptable[10,"xerror"])

which.min(OPT.POINT)

cv.tree31<-rpart(INCOME_C4~.,data = train.data3[,-1] ,method = "class", 

cp = 0.005876592, minsplit =5,xval =5, model=TRUE)         

prp(cv.tree31,type =0,extra = 1,under = T,split.font = 1,

varlen = -10,cex=.7,box.palette=c("mediumvioletred"))

cv.tree31.pred.train <- predict(cv.tree31,train.data3[,-1],type = "class")

confusionMatrix(cv.tree31.pred.train,train.data3[,-1]$INCOME_C4)

cv.tree31v<-rpart(INCOME_C4~., data = valid.data3[,-1] ,method = "class", cp =0.005876592 , minsplit =5,

xval =5, model=TRUE)

prp(cv.tree31v,type =0,extra = 1,under = T,split.font = 1,

varlen = -10,cex=.7,box.palette=c("mediumvioletred"))

#Validation

cv.tree31.pred.valid <- predict(cv.tree31v,valid.data3[,-1],type = "class")

confusionMatrix(cv.tree31.pred.valid,valid.data3[,-1]$INCOME_C4)