Interpretable Decision Trees Through MaxSAT

TL;DR

This paper introduces a MaxSAT-based method for constructing interpretable decision trees that achieve better accuracy and efficiency than traditional methods, enhancing the balance between interpretability and performance.

Contribution

It applies MaxSAT technology to efficiently compute minimum pure decision trees, outperforming existing approaches in accuracy and runtime.

Findings

MPDTs outperform sklearn decision trees in accuracy

The approach improves runtime efficiency

Enhanced interpretability with minimal tree size

Abstract

We present an approach to improve the accuracy-interpretability trade-off of Machine Learning (ML) Decision Trees (DTs). In particular, we apply Maximum Satisfiability technology to compute Minimum Pure DTs (MPDTs). We improve the runtime of previous approaches and, show that these MPDTs can outperform the accuracy of DTs generated with the ML framework sklearn.