LIBRE: Learning Interpretable Boolean Rule Ensembles

TL;DR

LIBRE is a novel ensemble method that learns interpretable Boolean rule classifiers using weak learners on feature subsets, balancing accuracy and interpretability effectively.

Contribution

It introduces a new ensemble approach combining weak Boolean rule learners with a simple union, enhancing interpretability without sacrificing accuracy.

Findings

LIBRE achieves competitive accuracy with black box models.

It provides highly interpretable Boolean rule ensembles.

LIBRE performs well on imbalanced datasets.

Abstract

We present a novel method - LIBRE - to learn an interpretable classifier, which materializes as a set of Boolean rules. LIBRE uses an ensemble of bottom-up weak learners operating on a random subset of features, which allows for the learning of rules that generalize well on unseen data even in imbalanced settings. Weak learners are combined with a simple union so that the final ensemble is also interpretable. Experimental results indicate that LIBRE efficiently strikes the right balance between prediction accuracy, which is competitive with black box methods, and interpretability, which is often superior to alternative methods from the literature.