Linear regression models, neural networks, and regression trees are the three methods that will be stacked here. We will require the caret
and caretEnsemble
packages to do this task. The stacked ensemble methods have been introduced in detail in Chapter 7, The General Ensemble Technique. First, we specify the control parameters for the training task, specify the list of algorithms, and create the stacked ensemble:
> control <- trainControl(method="repeatedcv", number=10, repeats=3, + savePredictions=TRUE, classProbs=TRUE) > algorithmList <- c('lm', 'rpart') > set.seed(12345) > Emodels <- caretList(HT_Formula, data=HT_Build, trControl=control, + methodList=algorithmList, + tuneList=list( + nnet=caretModelSpec(method='nnet', trace=FALSE, + linout=TRUE) + + ...