Machine Learning with R Cookbook - Second Edition

There's more...

In addition to using the boosting function in the gbm package, one can also use the mboost package to perform classifications with the gradient boosting method:

1. First, install and load the mboost package:

        > install.packages("mboost")
        > library(mboost) 

2. The mboost function only uses numeric responses; therefore, you should transform yes/no responses to numeric responses (0/1):

        > trainset$churn = ifelse(trainset$churn == "yes", 1, 0)

3. Also, you should remove nonnumerical attributes, such as voice_mail_plan and international_plan:

        > trainset$voice_mail_plan = NULL
        > trainset$international_plan = NULL  

4. We can then use mboost to train the classification model:

 > churn.mboost = mboost(churn ~ ., data=trainset, control ...