Granular Weighted Binning by Generalized Boosted Model

In the post https://statcompute.wordpress.com/2019/04/27/more-general-weighted-binning, I’ve shown how to do the weighted binning with the function wqtl_bin() by the iterative partitioning. However, the outcome from wqtl_bin() sometimes can be too coarse. The function wgbm_bin() (https://github.com/statcompute/MonotonicBinning/blob/master/code/wgbm_bin.R) leverages the idea of gbm() that implements the Generalized Boosted Model and generates more granular weighted binning outcomes.

Below is the demonstration showing the difference between wqtl_bin() and wgbm_bin() outcomes. Even with the same data, the wgbm_bin() function is able to generate a more granular binning result and 14% higher Information Value.

wtwoe