CFIRE: A General Method for Combining Local Explanations

TL;DR

The paper introduces CFIRE, a novel local-to-global XAI algorithm that derives global, faithful, compact, and complete rule-based explanations from local explanations for tabular data by integrating closed frequent itemset mining with axis-aligned box representations. It addresses the disagreement problem by allowing dynamic selection among multiple local explainers and tests robustness against the Rashomon effect across 50 high-performing black-box models and 14 benchmark datasets. CFIRE outputs a set of class-specific DNFs with a greedy set-cover step to ensure compactness, while closed frequent itemsets provide lossless compression and more precise feature combinations. Empirical results show CFIRE achieving superior faithfulness and completeness with significant runtime advantages over state-of-the-art baselines, and the paper discusses implications for cross-explainer robustness and potential future enhancements.

Abstract

We propose a novel eXplainable AI algorithm to compute faithful, easy-to-understand, and complete global decision rules from local explanations for tabular data by combining XAI methods with closed frequent itemset mining. Our method can be used with any local explainer that indicates which dimensions are important for a given sample for a given black-box decision. This property allows our algorithm to choose among different local explainers, addressing the disagreement problem, \ie the observation that no single explanation method consistently outperforms others across models and datasets. Unlike usual experimental methodology, our evaluation also accounts for the Rashomon effect in model explainability. To this end, we demonstrate the robustness of our approach in finding suitable rules for nearly all of the 700 black-box models we considered across 14 benchmark datasets. The results also show that our method exhibits improved runtime, high precision and F1-score while generating compact and complete rules.