weighted_vimportance_nestedrf: Weighted mean of variable importance for resampled RF learner

In messamat/globalIRmap: What the Package Does (One Line, Title Case)

Description

Compute mean and standard deviation of variable importance and mean of p-value across resampling instances (e.g. folds and repetitions) weighted by resampling prediction accuracy.

Usage

weighted_vimportance_nestedrf(rfresamp, pvalue = TRUE, pvalue_permutn = 100)

Arguments

rfresamp	the ResampleResult from a classification RF of type mlr_learners_classif.ranger or mlr_learners_classif.cforest
pvalue	(logical) Whether to compute p-values.
pvalue_permutn	(integer) number of permutations to use in p-value calculations

Details

See extract_impperf_nestedrf for more details on computations.

Value

A data.table with, when all logicals are TRUE, the following columns.

varnames

imp_wmean

instances

imp_wsd

resampling instances

pvalue_wmean

instances

Warning

Does not accept error measure for weighting

• could be amended to be based on 1-error

documentation to-do

Can add an example down the line, add source.

See Also

extract_impperf_nestedrf and ggvimp and benchmark_featsel

messamat/globalIRmap documentation built on July 4, 2021, 10:48 a.m.