An interpretable hypersphere information granule-based classifier for numeric data using axiomatic fuzzy set

The interpretability of classifiers focused on numerical data continues to pose significant challenges. This paper introduces an interpretable classifier rooted in axiomatic fuzzy set (AFS) theory and granular computing (GrC). The design of the proposed classifier consists of three stages: the construction of hypersphere information granules, the optimization of hypersphere information granules, and finding the semantic description of hypersphere information granules by applying AFS theory. The resulting classifier leverages hypersphere information granules to capture the data’s structural characteristics while facilitating the semantic interpretation of the overall structural landscape. Comprehensive evaluation across 17 diverse datasets demonstrates that the proposed classifier achieves better classification accuracy and interpretability.