# przyklad 6.2

options(OutDec=",")
library("ElemStatLearn")
library("mlbench")
library("rpart")
library("gbm")
library("ipred")

data(Vehicle)
m<-nrow(Vehicle)

veh.bag.blad<-matrix(ncol=10,nrow=100)
veh.bun.blad<-matrix(ncol=10,nrow=100)

comb.lda <- list(list(model=lda, predict=function(obj, newdata)
                  predict(obj, newdata)$x))


for (j in 1:100)
{

test <- sample(1:m, size = round(m/3), replace = FALSE) 
veh.ucz<-Vehicle[-test,]
veh.test<-Vehicle[test,]

y.test<-veh.test$Class

ks<-seq(5,50,by=5)

for (k in seq(along=ks)) 
{
print(ks[k])

lmod<-ks[k] # liczba modeli pojedynczych do klasyfikacji

# bagging
veh.bag<-bagging(Class~.,data=veh.ucz,nbagg=lmod)
Z.veh<-predict(veh.bag, newdata=veh.test, type="class",aggregation="majority")
veh.bag.blad[j,k]<-1-(sum(as.numeric(Z.veh==y.test))/length(y.test))

# bundling
veh.bun<- bagging(Class ~ .,data=veh.ucz,nbagg=lmod,comb=comb.lda) 
B.veh<-predict(veh.bun,newdata=veh.test, type="class",aggregation="majority")
veh.bun.blad[j,k]<-1-(sum(as.numeric(B.veh==y.test))/length(y.test))

}
}

veh.bag<-numeric(length=10)
for (k in 1:10) veh.bag[k]<-mean(veh.bag.blad[,k])

veh.bun<-numeric(length=10)
for (k in 1:10) veh.bun[k]<-mean(veh.bun.blad[,k])


# rysunek 6.3
plot(ks,veh.bag,type="n",ylim=c(min(veh.bun),max(veh.bag)),xlab="Liczba modeli bazowych",ylab="Błąd predykcji",main="Zbiór Vehicle")
lines(ks,veh.bag,type="l",lty=1)
lines(ks,veh.bun,type="l",lty=2)
points(ks,veh.bag,pch=19)
points(ks,veh.bun,pch=19)
legend("topright",lty=c(1,2),legend = c("Bagging", "Bundling"))