Advancements in Optimizing Fuzzy Grid Partition for Enhanced Rule Generation Performance: Algorithms, Interpretability, and Scalability

Abstract

This research focuses on optimizing fuzzy grid partitioning to enhance rule generation performance in fuzzy rule-based systems. A novel mathematical formulation is proposed, aiming to minimize the number of fuzzy grid cells while considering coverage, regularity, and overlap constraints. The study demonstrates the effectiveness of the approach through a case example in credit risk assessment. The optimized fuzzy grid partitioning scheme generates concise and interpretable fuzzy rules, improving the accuracy and interpretability of the rule-based system. The research highlights the significance of interpretability in rule-based systems and showcases the scalability and applicability of the approach across various domains. However, limitations include the lack of comprehensive comparisons, limited exploration of generalizability to different datasets, and the need for real-world implementation considerations. Nonetheless, this research provides valuable insights into optimizing fuzzy grid partitioning for rule generation and contributes to the advancement of fuzzy rule-based systems in decision support and problem-solving tasks. Future work should address the identified limitations and explore the practical implementation of the approach.