Principle:Interpretml Interpret APLR Regression
Metadata
| Field | Value |
|---|---|
| Sources | Interpretml_Interpret |
| Domains | Machine_Learning, Interpretability |
| Updated | 2026-02-07 |
Overview
Automatic Piecewise Linear Regression (APLR) provides inherently interpretable models by building piecewise linear functions with automatic term selection for both regression and classification tasks.
Description
APLR builds models composed of piecewise linear basis functions that are automatically selected during training. The algorithm produces models that are inherently interpretable because each term represents a simple piecewise linear relationship between a feature and the response. Both main effects and two-way interactions are supported. In the InterpretML framework, APLRRegressor handles regression tasks while APLRClassifier handles multi-class classification by internally using per-class logit APLR models.
Usage
Use APLR when you need an interpretable piecewise linear model with automatic feature and interaction selection. The models wrap the external aplr package and conform to the InterpretML explainer API, providing both global and local explanations.