setwd("E:/R/Data/")

wn.tmp <- read.csv("winequality-red.csv", header=T)
wn <- wn.tmp

# training dataset  testing dataset ɕB
qualityNames <- dimnames(table(wn$quality))[[1]]

testSampleRate <- 1/3

set.seed(789)      # e̊wДԍpĉB

wn.train <- NULL
wn.test <- NULL
for ( q in qualityNames ) {
  tmp <- subset(wn, wn$quality==as.numeric(q))
  tmpIndex <- sample.int(dim(tmp)[1])
  wn.test <- rbind( wn.test, tmp[tmpIndex[1:as.integer(testSampleRate*dim(tmp)[1])],] )
  wn.train <- rbind( wn.train, tmp[tmpIndex[as.integer(testSampleRate*dim(tmp)[1]+1):dim(tmp)[1] ],] )
}

head( wn.train[,c(1,12)] )
head( wn.test[,c(1,12)]  )

# t@Co

write.csv(wn.train, "winequality-red-train.csv", row.names=F)
write.csv(wn.test, "winequality-red-test.csv", row.names=F)

