bagg.default {TWIX}R Documentation

Predictions from TWIX's or Bagging Trees

Description

Prediction of a new observation based on multiple trees.

Usage

## Default S3 method:
bagg(object,newdata=NULL,sq=1:length(object$trees),
                aggregation="weighted",...)
## S3 method for class 'TWIX':
bagg(object,...)
## S3 method for class 'bootTWIX':
bagg(object,...)

Arguments

object object of classes TWIX or bootTWIX.
newdata a data frame of new observations.
sq Integer vector indicating for which trees predictions are required.
aggregation character string specifying how to aggregate. There are two ways to aggregate the TWIX trees. One of them is the class majority voting (aggregation="majority") and another method is the weighted aggregation (aggregation="weighted").
... additional arguments.

See Also

TWIX, predict.TWIX, bootTWIX

Examples

    library(ElemStatLearn)
    data(SAheart)

    ### response variable must be a factor
    SAheart$chd <- factor(SAheart$chd) 

    ### test and train data
    ###
    set.seed(1234)
    icv <- sample(nrow(SAheart),nrow(SAheart)/3)
    itr <- setdiff(1:nrow(SAheart),icv)
    train <- SAheart[itr,]
    test <- SAheart[icv,]

    ### TWIX Ensemble
    M1 <- TWIX(chd~.,data=train,topN=c(4,3),topn.method="single")

    ### Bagging with greedy decision trees as base classifier
    M2 <- bootTWIX(chd~.,data=train,nbagg=30)

    ### Bagging with the p-value adjusted classification trees as base classifier
    M3 <- bootTWIX(chd~.,data=train,nbagg=30,splitf="p-adj",Devmin=0.01)

    preda <- bagg(M1,test,sq=1:length(M1$trees),aggregation="majority")
    predb <- bagg(M1,test,sq=1:length(M1$trees),aggregation="weighted")
    pred1 <- predict(M2,test,sq=1:length(M2$trees))
    pred2 <- predict(M3,test,sq=1:length(M3$trees))

    ###
    ### CCR's
  
    sum(preda == test$chd)/nrow(test)
    sum(predb == test$chd)/nrow(test)
    sum(pred1 == test$chd)/nrow(test)
    sum(pred2 == test$chd)/nrow(test)

[Package TWIX version 0.2.6 Index]