Search code examples
rrandom-forestfeature-selectiontidymodels

How to evaluate joint importance of two features in a model (random forest) using R package such as VIP or DALEXtra?

I know to use these packages (VIP etc) with tidymodels to evaluate individual feature contribution/importance for a model such as random forest. But I'd like to know to evaluate a combined or joint importance for two or more features? Probably I cannot just simply add importance scores of these features together?

Thank you.

-Xiaokuan

Solution

  • You can use variable_groups argument in DALEX::model_parts function.