Variable Importance Measures for Multivariate Random Forests

Calculates two sets of post-hoc variable importance measures for multivariate random forests. The first set of variable importance measures are given by the sum of mean split improvements for splits defined by feature j measured on user-defined examples (i.e., training or testing samples). The second set of importance measures are calculated on a per-outcome variable basis as the sum of mean absolute difference of node values for each split defined by feature j measured on user-defined examples (i.e., training or testing samples). The user can optionally threshold both sets of importance measures to include only splits that are statistically significant as measured using an F-test.

MulvariateRandomForestVarImp

The goal of MulvariateRandomForestVarImp package is to calculates post-hoc variable importance measures for multivariate random forests. These are given by split improvement for splits defined by feature j as measured using user-defined (i.e. training or test) examples. Importance measures can also be calculated on a per-outcome variable basis using the change in predictions for each split. Both measures can be optionally thresholded to include only splits that produce statistically significant changes as measured by an F-test.

Installation

You can install the released version of VIM from CRAN with:

install.packages("MulvariateRandomForestVarImp")

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("Megatvini/VIM")

Example

This is a basic example which shows you how use the package:

library(MulvariateRandomForestVarImp)
## basic example code
set.seed(49)

X <- matrix(runif(50*5), 50, 5)
Y <- matrix(runif(50*2), 50, 2)

split_improvement_importance <- MeanSplitImprovement(X, Y)
split_improvement_importance
#> [1] 0.8066173 2.8909635 3.4591123 0.6227943 0.5138745

mean_outccome_diff_importance <- MeanOutcomeDifference(X, Y)
mean_outccome_diff_importance
#>           [,1]      [,2]
#> [1,] 0.2458139 0.3182474
#> [2,] 0.2712269 0.2915053
#> [3,] 0.2125802 0.2023291
#> [4,] 0.2819759 0.2519035
#> [5,] 0.1238451 0.1958629