# [R] ensemble methods

Bert Gunter gunter.berton at gene.com
Mon Nov 11 19:49:21 CET 2013

```See the R randomForest package.

This already does ensemble classification and regression.

-- Bert

On Mon, Nov 11, 2013 at 10:04 AM, Iut Tri Utami <triutami.iut at gmail.com> wrote:
> Dear Mr/Mrs
>
> I am Iut, student of graduate student in Bogor Agriculture Institur
> I read a book on ensemble methods in data mining by Seni and Elder and find
> I am confused how to call these functions and and how to agregate it with
> I think there is missing code in here.What if the function is replaced with
> SVM?
>
> Example :
> genPredictors <- function(seed = 123, N = 30) {
> # Load package with random number generation
> # for the multivariate normal distribution
> library(mnormt)
> # 5 "features" each having a "standard" Normal
> # distribution with pairwise correlation 0.95
> Rho <- matrix(c(1,.95,.95,.95,.95,
> + .95, 1,.95,.95,.95,
> + .95,.95,1,.95,.95,
> + .95,.95,.95,1,.95,
> + .95,.95,.95,.95,1), 5, 5)
> mu <- c(rep(0,5))
> set.seed(seed);
> x <- rmnorm(N, mu, Rho)
> colnames(x) <- c("x1", "x2", "x3", "x4", "x5")
> return(x)
> }
> genTarget <- function(x, N, seed = 123) {
> # Response Y is generated according to:
> # Pr(Y = 1 | x1 <= 0.5) = 0.2,
> # Pr(Y = 1 | x1 > 0.5) = 0.8
> y <- c(rep(-1, N))
> set.seed(seed);
> for (i in 1:N) {
> if ( x[i,1] <= 0.5 ) {
> if ( runif(1) <= 0.2 ) {
> y[i] <- 1
> } else {
> y[i] <- 0
> }
> } else {
> if ( runif(1) <= 0.8 ) {
> y[i] <- 1
> } else {
> y[i] <- 0
> }
> }
> }
> return(y)
> }
> genBStrapSamp <- function(seed = 123, N = 200, Size = 30) {
> set.seed(seed)
> sampleList <- vector(mode = "list", length = N)
> for (i in 1:N) {
> sampleList[[i]] <- sample(1:Size, replace=TRUE)
> }
> return(sampleList)
> }
> fitBStrapTrees <- function(data, sampleList, N) {
> treeList <- vector(mode = "list", length = N)
> for (i in 1:N) {
> tree.params=list(minsplit = 4, minbucket = 2, maxdepth = 7)
> treeList[[i]] <- fitClassTree(data[sampleList[[i]],],
> tree.params)
> }
> return(treeList)
> }
> fitClassTree <- function(x, params, w = NULL,
> seed = 123) {
> library(rpart)
> set.seed(seed)
> tree <- rpart(y ~ ., method = "class",
> data = x, weights = w, cp = 0,
> minsplit = params.minsplit,
> minbucket = params.minbucket,
> maxdepth = params.maxdepth)
> return(tree)
> }
>
> thankyou very much
>
> best regard,
>
> Iut
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> and provide commented, minimal, self-contained, reproducible code.

--

Bert Gunter
Genentech Nonclinical Biostatistics

(650) 467-7374

```