Desktop Survival Guide
by Graham Williams
A number of R packages implement boosting. The caTools package provides the LogitBoost function which is perhaps the simplest to use, and is an efficient implementation for large datasets. The boost package provides the adaboost function as well as logitboost, and relies on rpart for building the models, and is less efficient. The gbm package is the more sophisticated of the packages and implements the more general Generalise Boosted Regression Models. We will illustrate boosting with the gbm package.
We start our examples though with a step through of the process using just rpart.