Dynamic Interpretability for Model Comparison via Decision Rules

TL;DR

This paper introduces DeltaXplainer, a model-agnostic method that generates rule-based explanations to compare differences between two classifiers, aiding model selection and monitoring in real-world scenarios.

Contribution

The paper presents DeltaXplainer, a novel approach for explaining differences between models, addressing a gap in existing XAI methods focused on single models.

Findings

DeltaXplainer effectively explains differences in synthetic datasets.

It successfully identifies concept drift in real-world datasets.

The method is applicable across various model comparison scenarios.

Abstract

Explainable AI (XAI) methods have mostly been built to investigate and shed light on single machine learning models and are not designed to capture and explain differences between multiple models effectively. This paper addresses the challenge of understanding and explaining differences between machine learning models, which is crucial for model selection, monitoring and lifecycle management in real-world applications. We propose DeltaXplainer, a model-agnostic method for generating rule-based explanations describing the differences between two binary classifiers. To assess the effectiveness of DeltaXplainer, we conduct experiments on synthetic and real-world datasets, covering various model comparison scenarios involving different types of concept drift.