I am being asked what the units are for variable importance using the varimp() function for a caret package random forest model. My best answer based on previous posts is % but is that accurate? Would %MSE be a better interpretation, something else entirely or is it unitless and just showing relative magnitude? The caret documentation does not say and previous posts are not definitive on this.
Asked
Active
Viewed 79 times
0
-
you can find the answer in the following topic: Why does random forest variable importance not sum to 100 – Eula Carrara Oct 03 '23 at 13:38