# Lab: Decision Trees


## Fitting Classification Trees

###
library(tree)
###
library(ISLR2)
attach(Carseats)
High <- factor(ifelse(Sales <= 8, "No", "Yes"))
###
Carseats <- data.frame(Carseats, High)
###
tree.carseats <- tree(High ~ . - Sales, Carseats)
?tree
## default je split=deviance koristi entropiju kao mjeru necistoce u slucaju klasifikacije
###
summary(tree.carseats)
### devijanca je u nasoj notaciji tocno n*C(T) (vidi (7.12))
plot(tree.carseats)
text(tree.carseats, pretty = 0, cex=0.7)
### prva podjela je shelveLoc= Bad ili Medium, te ShelveLoc=Good
tree.carseats
###
set.seed(2)
train <- sample(1:nrow(Carseats), 200)
Carseats.test <- Carseats[-train, ]
High.test <- High[-train]
tree.carseats <- tree(High ~ . - Sales, Carseats,
                      subset = train)
tree.pred <- predict(tree.carseats, Carseats.test,
                     type = "class")
table(tree.pred, High.test)
(104 + 50) / 200
### ovo je 1- testna greska
set.seed(7)
cv.carseats <- cv.tree(tree.carseats, FUN = prune.misclass) # za obrezivanje koristimo 
                                                            # 0-1 mjeru necistoce
cv.carseats
### dev je sada ukupan broj primjera koji su bili krivo klasificirani u CV metodi
plot(cv.carseats, type="b")

###
prune.carseats <- prune.misclass(tree.carseats, best = 9)
plot(prune.carseats)
text(prune.carseats, pretty = 0)
###
tree.pred <- predict(prune.carseats, Carseats.test,
                     type = "class")
table(tree.pred, High.test)
(97 + 58) / 200
### nije veliko povecanje u preciznosti, ali dobili smo manje (dakle, intepretabilnije) stablo