library(MASS)
data(iris)
head(iris)

Species <- iris$Species
levels(Species)
g <- length(levels(Species))
p <- 4
n <- dim(X)[1]
X <- iris[,1:p]


lda.obj <- lda(X,Species)
lda.obj

q <- min(2,p)
m <- cbind(1:q,lda.obj$svd,100*lda.obj$svd/sum(lda.obj$svd),100*cumsum(lda.obj$svd)/sum(lda.obj$svd))
colnames(m) <- c("e-pair","eigenvalue","%acc't by","% cumulative acc't by")
options(digits = 3)
m


LD.f <- lda.obj$scaling
X.m <- matrix(as.numeric(unlist(X)),n,p)
scores <- X.m%*%LD.f
plot(scores[,1],scores[,2],type="n")

for (i in 1:g){
	I <- Species == levels(Species)[i]
	points(scores[I,1],scores[I,2],col=i)}
##########################################################
head(fgl)
lda.obj <- lda(type~.,data=fgl)

n <- dim(fgl)[1]
p <- dim(fgl)[2]-1
q <- min(5,p)

m <- cbind(1:q,lda.obj$svd,100*lda.obj$svd/sum(lda.obj$svd),100*cumsum(lda.obj$svd)/sum(lda.obj$svd))
colnames(m) <- c("e-pair","eigenvalue","%acc't by","% cumulative acc't by")
options(digits = 3)
m

LD.f <- lda.obj$scaling
X.m <- matrix(as.numeric(unlist(fgl[,1:9])),n,p)

scores <- X.m%*%LD.f
plot(scores[,1],scores[,2],type="n")
g <- length(levels(fgl$type))
for (i in 1:g){
	I <- fgl$type == levels(fgl$type)[i]
	points(scores[I,1],scores[I,2],col=i)}


method <- 1 # random forest

libary(randomForest)
method <- 2 #lda

yX <- data.frame(fgl$type,X.m)
r <- 50
nt <- floor(0.9*n)
for (i in 1:r){

	s <- sample(1:n,size=nt)
	u <- rep(0,nt)
	ct <- 1
	for (j in 1:n){
		if (sum(s==j)==0) {
			u[ct] <- j
			ct <- ct + 1}
	}

	if (method ==1){
		rf.obj <- randomForest(x=X.m,y=fgl$type,
			xtest=X.m[u,],ytest=fgl$type[u])
		predictions <- rf.obj$predicted}

	if (method==2){
		lda.obj <-  lda(fgl$type~X.m,CV=F,subset=s)  #,priors=rep(1/6,6)
		predictions <- predict(lda.obj,newdata=yX)$class
	}
	if (i==1)	save.m <- cbind(fgl$type[u],predictions[u])
	if (i!=1)	save.m <- rbind(save.m,cbind(fgl$type[u],predictions[u]))
	print(c("Loop ",i))
}

y <- save.m[,1]
predictions <- save.m[,2]
tabl <- table(y,predictions)

options(digits=3)

100*diag(tabl/rowSums(tabl))
rowSums(tabl)
100*sum(diag(tabl))/sum(tabl)

tabl2 <- rbind(cbind(tabl,rowSums(tabl)),c(colSums(tabl),sum(sum(tabl))))
rownames(tabl2) <- c(rownames(tabl),"Total")
colnames(tabl2) <- c(colnames(tabl),"Total")
"Rows = actual       columns = predicted"
tabl2