Using Model-Based Trees with Boosting to Fit Low-Order Functional ANOVA Models

TL;DR

This paper introduces GAMI-Tree, a novel algorithm combining model-based trees and interaction filtering to improve the performance and interpretability of low-order functional ANOVA models in machine learning.

Contribution

GAMI-Tree integrates model-based trees with a new interaction filtering method, enhancing predictive accuracy and interpretability over existing methods like EBM and GAMI-Net.

Findings

GAMI-Tree outperforms EBM and GAMI-Net in predictive accuracy.

The algorithm converges faster and requires less tuning.

GAMI-Tree provides more interpretable models with hierarchical orthogonality.

Abstract

Low-order functional ANOVA (fANOVA) models have been rediscovered in the machine learning (ML) community under the guise of inherently interpretable machine learning. Explainable Boosting Machines or EBM (Lou et al. 2013) and GAMI-Net (Yang et al. 2021) are two recently proposed ML algorithms for fitting functional main effects and second-order interactions. We propose a new algorithm, called GAMI-Tree, that is similar to EBM, but has a number of features that lead to better performance. It uses model-based trees as base learners and incorporates a new interaction filtering method that is better at capturing the underlying interactions. In addition, our iterative training method converges to a model with better predictive performance, and the embedded purification ensures that interactions are hierarchically orthogonal to main effects. The algorithm does not need extensive tuning, and our implementation is fast and efficient. We use simulated and real datasets to compare the performance and interpretability of GAMI-Tree with EBM and GAMI-Net.