This is the source code for the maidrr package, which is currently under development.
Installation
To install maidrr from GitHub you will need devtools:
install.packages('devtools') devtools::install_github('henckr/maidrr')
Overview
The goal of maidrr is to aid you in the development of a Model-Agnostic Interpretable Data-driven suRRogate for your black box algorithm of choice. In short, these are the steps in the procedure:
- Partial dependencies (PDs) are used to obtain model insights from the black box in the form of feature effects.
- Those effects are used to group values/levels within a feature in an optimal data-driven way, while performing built-in feature selection.
- An interpretable GLM surrogate is fit to the segmented features. Meaningful interactions can be included if desired.