install.packages('rpart')

library(rpart)
par(mfrow = c(1,1))
library(rpart)
data(kyphosis)
head(kyphosis)
fit1 <- rpart(Kyphosis ~ ., data = kyphosis,
              parms=list(split = "information"))
fit1
summary(fit1)
plot(fit1)
text(fit1, use.n = TRUE)
predict(fit1, kyphosis,type = "prob")   # class probabilities (default)
table(predict(fit1, kyphosis,type = "class"),kyphosis$Kyphosis)

# better than LDA? 
library(MASS)
fit.lda <- lda(Kyphosis ~ ., data = kyphosis)
fit.lda # show results
table(predict(fit.lda)$class, kyphosis$Kyphosis)

# compare various options
par(mfrow = c(1,3))
fit <- rpart(Kyphosis ~ ., data = kyphosis,
              parms=list(split = "information"),
              control = rpart.control(cp = 0.03))
plot(fit)
text(fit, use.n = TRUE)
fit <- rpart(Kyphosis ~ ., data = kyphosis,
              parms=list(split = "gini"))
plot(fit)
text(fit, use.n = TRUE)
fit <- rpart(Kyphosis ~ ., data = kyphosis,
              parms=list(split = "gini"),
              control = rpart.control(cp = 0.05))
plot(fit)
text(fit, use.n = TRUE)


par(mfrow = c(1,1))
# iris data (3-group classification)
sub <- c(sample(1:50, 25), sample(51:100, 25), sample(101:150, 25))
fit <- rpart(Species ~ ., data = iris, subset = sub)
fit.pruned.tree <-prune(fit, cp = 0.1)
fit.pruned.tree 
table(predict(fit.pruned.tree , iris[-sub,], type = "class"), iris[-sub, "Species"])

plot(fit)
text(fit, use.n = TRUE,cex = 0.7)


## regression

library(MASS)
data(Boston)
head(Boston)
lm.out<-lm(medv ~ ptratio, data = Boston)
plot(medv ~ ptratio, data = Boston)
abline(lm.out, col="red")

fit <- rpart(medv ~ ptratio, data = Boston)
fit.pruned.tree <-prune(fit, cp = 0.05) 
plot(fit.pruned.tree)
text(fit.pruned.tree, use.n = TRUE)
fit.pruned.tree
plot(medv ~ ptratio, data = Boston)
ord = order(Boston$ptratio)
lines(Boston$ptratio[ord],predict(fit.pruned.tree,Boston[ord,]))

fit <- rpart(medv ~ ptratio, data = Boston)
fit.pruned.tree <-prune(fit, cp = 0.05) 
plot(fit.pruned.tree)
text(fit.pruned.tree, use.n = TRUE)
fit.pruned.tree
plot(medv ~ ptratio, data = Boston)
ord = order(Boston$ptratio)
lines(Boston$ptratio[ord],predict(fit.pruned.tree,Boston[ord,]))
 

fit <- rpart(medv ~ ptratio + nox, data = Boston)
fit.pruned.tree <-prune(fit, cp = 0.01) 
plot(fit.pruned.tree)
text(fit.pruned.tree, use.n = TRUE)
fit.pruned.tree