credit_default <- read.csv("credit.csv",header=TRUE, sep=",")

credit_default <- credit_default[which(credit_default$EDUCATION %in% c(1, 2, 3, 4)), ]
credit_default <- credit_default[which(credit_default$MARRIAGE != 0), ]
credit_default <- credit_default[, -1]

credit_default[, c(2:4, 6:11, 24)] <- lapply(credit_default[, c(2:4, 6:11, 24)], factor)
str(credit_default)

attach(credit_default)

library(MASS)
library(tree) ##to fit trees
library(randomForest) ##for random forests (and bagging)
library(gbm)

#regression
set.seed(1)
sample.data<-sample.int(nrow(credit_default), floor(.50*nrow(credit_default)), replace = F)
train<-credit_default[sample.data, ]
test<-credit_default[-sample.data, ]

tree.class.train<-tree(LIMIT_BAL~default.payment.next.month+SEX+EDUCATION+MARRIAGE+PAY_0+PAY_2+PAY_3+PAY_4+PAY_5+PAY_6+BILL_AMT1+BILL_AMT2+BILL_AMT3+BILL_AMT4+BILL_AMT5+BILL_AMT6+PAY_AMT1+PAY_AMT2+PAY_AMT3+PAY_AMT4+PAY_AMT5+PAY_AMT6, data=train)
plot(tree.class.train)
text(tree.class.train, cex=0.75, pretty=0)

summary(tree.class.train)

# pruning

set.seed(11)
cv.credit<-cv.tree(tree.class.train, K=10)
cv.credit

plot(cv.credit$size, cv.credit$dev, type='b', ylab="Deviance", xlab="Size")

trees.num<-cv.credit$size[which.min(cv.credit$dev)]
print(trees.num)

prune.credit<- prune.tree(tree.class.train, best = trees.num)
plot(prune.credit)
text(prune.credit, cex = 0.75, pretty = 0)

yhat<-predict(prune.credit, newdata=test)
credit.test<-test[,"LIMIT_BAL"]
mse.tree<-mean((credit.test-yhat)^2)
print(mse.tree)


#classification
tree.class.train<-tree(default.payment.next.month~LIMIT_BAL+SEX+EDUCATION+MARRIAGE+PAY_0+PAY_2+PAY_3+PAY_4+PAY_5+PAY_6+BILL_AMT1+BILL_AMT2+BILL_AMT3+BILL_AMT4+BILL_AMT5+BILL_AMT6+PAY_AMT1+PAY_AMT2+PAY_AMT3+PAY_AMT4+PAY_AMT5+PAY_AMT6)
plot(tree.class.train)
text(tree.class.train, cex=0.75, pretty=0)
summary(tree.class.train)