library(plyr)
CVgroup <- function(k,datasize,seed){
 cvlist <- list()
 set.seed(seed)
 n <- rep(1:k,ceiling(datasize/k))[1:datasize]  #將數(shù)據(jù)分成K份，并生成的完成數(shù)據(jù)集n
 temp <- sample(n,datasize)  #把n打亂
 x <- 1:k
 dataseq <- 1:datasize
 cvlist <- lapply(x,function(x) dataseq[temp==x]) #dataseq中隨機(jī)生成k個(gè)隨機(jī)有序數(shù)據(jù)列
 return(cvlist)
}

k <- 10
datasize <- nrow(iris)
cvlist <- CVgroup(k = k,datasize = datasize,seed = 1206)
cvlist

data <- iris
pred <- data.frame()  #存儲(chǔ)預(yù)測(cè)結(jié)果
library(plyr)
library(randomForest)
m <- seq(60,500,by = 20) #如果數(shù)據(jù)量大盡量間隔大點(diǎn)，間隔過(guò)小沒(méi)有實(shí)際意義
for(j in m){  #j指的是隨機(jī)森林的數(shù)量
 progress.bar <- create_progress_bar("text") #plyr包中的create_progress_bar函數(shù)創(chuàng)建一個(gè)進(jìn)度條，
 progress.bar$init(k)  #設(shè)置上面的任務(wù)數(shù)，幾折就是幾個(gè)任務(wù)
 for (i in 1:k){
  train <- data[-cvlist[[i]],] #剛才通過(guò)cvgroup生成的函數(shù)
  test <- data[cvlist[[i]],]
  model <-randomForest(Sepal.Length~.,data = train,ntree = j)  #建模，ntree=j 指的樹(shù)數(shù)
  prediction <- predict(model,subset(test,select = -Sepal.Length))  #預(yù)測(cè)
  randomtree <- rep(j,length(prediction))  #隨機(jī)森林樹(shù)的數(shù)量
  kcross <- rep(i,length(prediction))  #i是第幾次循環(huán)交叉，共K次
  temp <- data.frame(cbind(subset(test,select = Sepal.Length),prediction,randomtree,kcross))#真實(shí)值、預(yù)測(cè)值、隨機(jī)森林樹(shù)數(shù)、預(yù)測(cè)組編號(hào)捆綁在一起組成新的數(shù)據(jù)框tenp
  pred <- rbind(pred,temp)  #temp按行和pred合并
  print(paste("隨機(jī)森林：",j)) #循環(huán)至樹(shù)數(shù)j的隨機(jī)森林模型
  progress.bar$step() #輸出進(jìn)度條。告知完成了這個(gè)任務(wù)的百分之幾
 }
}

data <- iris
library(plyr)
library(randomForest)
k = 10
j <- seq(10,10000,by = 20)  #j樹(shù)的數(shù)量
i <- 1:k  #K折
i <- rep(i,times = length(j))
j <- rep(j,each = k)  #多少折，each多少
x <- cbind(i,j)
cvtest <- function(i,j){
 train <- data[-cvlist[[i]],]
 test <- data[cvlist[[i]],]
 model <- randomForest(Sepal.Length~.,data = train,ntree = j)
 prediction <- predict(model,subset(test,select = -Sepal.Length))
 temp <- data.frame(cbind(subset(test,select = Sepal.Length),prediction))
}

system.time(pred <- mdply(x,cvtest))