決定木のプログラム

決定木のプログラムメモ

library(rpart)
library(ROCR)

Data <- read.csv("...", as.is=T)

head(Data)
names(Data)
dim(Data)

tree01 <- rpart(Flag ~ ., data=Data, control=rpart.control(cp = 0.05))

par(mar=c(2, 2, 2, 2))
plot(tree01)
text(tree01, use.n = TRUE)
tree01

sink("...")
  tree01
sink()


#  木の性能チェック
pred <- 1 - predict(tree01)[1:nrow(Data)]
table(pred, Data$Flag)

prediction <- prediction(pred, factor(Data$Flag))
roc <- performance(prediction, "tpr", "fpr")


par(mar=c(3, 3, 3, 3))
plot(roc)

auc <- performance(prediction, "auc")
as.numeric(auc@y.values)