R菜鳥想請問R該怎跑10 fold

r cross-validation sk-learning

monkey9320 2017-05-10 21:52:43 ‧ 6598 瀏覽

分享至

這是我目前用R寫決策樹的部分，但卻不知如何進行10-fold來相互驗證，想請問各方高手該如何撰寫~

library(rpart)
library(rpart.plot)

dat = read.csv("test.csv")

dat=dat[,c(-12,-13,-15,-14)]

require(rpart)
set.seed(5566)


train=dat[(as.numeric(0.1*nrow(dat))+1):as.numeric(0.9*nrow(dat)),]
test= dat[(as.numeric(0.9*nrow(dat))+1):as.numeric(1*nrow(dat)),]
cart.model<- rpart(cluster ~. ,dat=train)
require(rpart.plot) 
prp(cart.model,fallen.leaves=TRUE,shadow.col="gray")
pred = predict(cart.model, newdat=test, type="class")
table(real=test$cluster, predict=pred)
confus.matrix <- table(real=test$cluster, predict=pred)
sum(diag(confus.matrix))/sum(confus.matrix)