varImpGroup: A grouped variable importance with Random Forests

In RFgroove: Importance Measure and Selection for Groups of Variables with Random Forests

Description

A permutation variable importance for groups of variables with Random Forests.

Usage

varImpGroup(object, xdata, ngroups = length(nvarGroup), nvarGroup,
            idxGroup, groupsNames = names(nvarGroup), 
            normalize = (length(unique(nvarGroup)) != 1))

Arguments

object	A randomForest object.
xdata	The input data.
ngroups	The number of groups.
nvarGroup	The vector of the number of variables in each group.
idxGroup	A list of size ‘ngroups’ containing the indexes of each group starting from 0.
groupsNames	The group names.
normalize	Should the normalized grouped importance measure be computed.

Value

An object of class ‘importance’ which is a vector of the importance for each group.

Author(s)

Baptiste Gregorutti

References

Gregorutti, B., Michel, B. and Saint Pierre, P. (2015). Grouped variable importance with random forests and application to multiple functional data analysis, Computational Statistics and Data Analysis 90, 15-35.

See Also

selectGroup,selectLevel,selectFunctional,plot.importance

Examples

  data(toyClassif)
  attach(toyClassif)

  rf <- randomForest(x=X,y=Y,keep.forest=TRUE, keep.inbag=TRUE, ntree=500)
  ngroups <- 3
  nvarGroup <- c(4,3,6)
  idxGroup <- list(c(0,1,2,5), c(2,4,5), c(0,1,5,6,7,8))
  grImp <- varImpGroup(rf, X, ngroups, nvarGroup, idxGroup, NULL, normalize=FALSE )
  cat("Group importance\n", grImp, "\n")

  detach(toyClassif)

Example output

Loading required package: randomForest
randomForest 4.6-12
Type rfNews() to see new features/changes/bug fixes.
Loading required package: wmtsa
Loading required package: fda
Loading required package: splines
Loading required package: Matrix

Attaching package: 'fda'

The following object is masked from 'package:graphics':

    matplot

Group importance
 0.2701206 0.1197495 0.1741828 

RFgroove documentation built on May 2, 2019, 8:31 a.m.