Interpretable Fuzzy Rule-Based Regression Extension for Ex-Fuzzy Library

2026-07-22

AI Summary

This work proposes an interpretable regression extension to the Ex-Fuzzy library based on Mamdani inference, specifically designed for safety-critical and regulated domains where high-accuracy models are often hindered by poor interpretability. The approach initializes linguistic variables via a target-aware Fuzzy C-Means clustering in an augmented input–output space, emphasizing feature regions most sensitive to the output. By integrating scalar consequent rule learning with output-oriented fuzzy partitions, it constructs a compact and human-readable rule base. Empirical evaluation on ten KEEL regression datasets demonstrates that the method achieves an average coefficient of determination (R²) of 0.86 using only 10–15 rules—significantly outperforming uniform partitioning and matching the predictive performance of black-box models such as random forests—thus effectively balancing interpretability and accuracy.

Abstract

Machine learning models achieve high predictive accuracy in regression tasks, but their deployment in safety-critical and regulated domains requires interpretability. While fuzzy rule-based systems offer transparent, linguistically explicit interpretable models, Mamdani-style fuzzy regression remains underrepresented in modern machine learning software libraries. This paper presents an interpretable regression extension for the Ex-Fuzzy library, enabling Mamdani fuzzy inference with scalar consequents learned directly from data. For this, a target-aware partition initialisation strategy based on Fuzzy C-Means clustering is introduced, in which linguistic variables are derived from an augmented input-output space to emphasise output-relevant regions of the feature space. The proposed extension is evaluated on ten regression datasets from the KEEL repository, comparing Gaussian and trapezoidal partition strategies against standard baselines including linear regression, multilayer perceptron, and random forests. Experimental results show that Gaussian partitions consistently outperform uniform trapezoidal partitions, achieving a mean coefficient of determination of approximately 0.86 while producing compact rule bases of 10-15 human-readable rules. The proposed implementation provides a transparent and competitive alternative to black-box regression models, supporting practical interpretability with competitive predictive performance.

Problem

• interpretable regression
• fuzzy rule-based systems
• Mamdani fuzzy inference
• machine learning interpretability
• safety-critical domains

Innovation

• Mamdani fuzzy regression
• interpretable machine learning
• Fuzzy C-Means clustering
• target-aware partitioning
• Ex-Fuzzy library