tdmPreGroupLevels: Group the levels of factor variable in 'dset[,colname]'.

In TDMR: Tuned Data Mining in R

Description

This function reduces the number of levels for factor variables with too many levels. It counts the cases in each level and orders them decreasingly. It binds the least frequent levels together in a new level "OTHER" such that the remaining untouched levels have more than opts$PRE.Xpgroup percent of all cases. OR it binds the levels with least cases together in "OTHER" such that the total number of new levels is opts$PRE.MaxLevel. From these two choices for "OTHER" take the one which binds more variables in column "OTHER".

Usage

tdmPreGroupLevels(dset, colname, opts)

Arguments

dset	data frame
colname	name of column to be re-grouped
opts	list, here we need PRE.Xpgroup [0.99] PRE.MaxLevel [32] (32 is the maximum number of levels allowed for randomForest)

Value

dset, a data frame with dset[,colname] re-grouped

TDMR documentation built on March 3, 2020, 1:06 a.m.