1

请参阅下面的循环 - 它调用 plot k(k-1)/2 次,但实际上没有生成任何图。但是,如果我更改代码以手动调用图(例如 plot(my_tree,c(1,2 ),..),地块已生成。)

my_tree 是一个 GBM 对象。请参阅下面的完整代码

#there is nothing wrong with this code, but yet it does not work
for (i in my_tree$var.names) {
        for (j in my_tree$var.names) {
                if (i < j) plot(my_tree, c(i, j), n.trees=best.iter)
        }
}

==== 下面的完整程序

pdf("demo.pdf")

N <- 100000
alldata <- data.frame();
train <- factor( ifelse(runif(N)<0.5, 'T', 'V'))
X1 <- rnorm(N);
X2 <- rnorm(N);
X3 <- runif(N);
X4 <- rpois(N, lambda=4);
linear.pred <- -3 + 0.25*X1 + 0.125*X2 - X3 + X4**abs(X1)
temp <- binomial()
y <- rbinom(N, 1, p=temp$linkinv(linear.pred))

alldata <- data.frame(train,X1,X2,X3,X4,y)
rm(train,X1,X2,X3,X4,y,linear.pred)

train <- alldata[alldata$train=='T',]

library(gbm)
my_tree<-gbm(y ~ X1 + X2 + X3 + X4,
    distribution="bernoulli",
    data=train,
    train.fraction=0.5,
    interaction.depth=8,
    n.trees=300,
    shrinkage=0.1,
    verbose=TRUE)

best.iter <- gbm.perf(my_tree, method="test")
print(best.iter)

summary(my_tree, ntrees=best.iter)

# make one and two ways
for (i in my_tree$var.names) {
        plot(my_tree, i, best.iter)      
}

#there is nothing wrong with this code, but yet it does not work
for (i in my_tree$var.names) {
        for (j in my_tree$var.names) {
                if (i < j) plot(my_tree, c(i, j), n.trees=best.iter)
        }
}
4

1 回答 1

2

如果您查看帮助页面,plot.gbm它提到使用了格子图,那么这很可能是 FAQ 7.22 的一个案例。尝试“打印”这些图。

于 2012-10-04T16:34:22.497 回答