SplitWise Regression: Stepwise Modeling with Adaptive Dummy Encoding

TL;DR

SplitWise is a new regression framework that adaptively transforms numeric predictors into binary features using shallow decision trees, improving model fit while maintaining interpretability, and outperforming traditional methods in various datasets.

Contribution

It introduces an adaptive dummy encoding method integrated with stepwise regression, enhancing nonlinear modeling without losing interpretability.

Findings

Produces more parsimonious models than traditional methods

Improves model generalization on real-world datasets

Easily implemented as an R package

Abstract

Capturing nonlinear relationships without sacrificing interpretability remains a persistent challenge in regression modeling. We introduce SplitWise, a novel framework that enhances stepwise regression. It adaptively transforms numeric predictors into threshold-based binary features using shallow decision trees, but only when such transformations improve model fit, as assessed by the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC). This approach preserves the transparency of linear models while flexibly capturing nonlinear effects. Implemented as a user-friendly R package, SplitWise is evaluated on both synthetic and real-world datasets. The results show that it consistently produces more parsimonious and generalizable models than traditional stepwise and penalized regression techniques.