Handling interaction in fuzzy production rule reasoning

Area of Science:

• Artificial Intelligence
• Fuzzy Logic
• Computational Intelligence

Background:

• Approximate reasoning with fuzzy production rules often suffers from rule interaction, where rules with the same consequent negatively impact accuracy.
• The conventional weighted average model fails to adequately address these interactions in real-world applications.

Purpose of the Study:

• To propose a new method for handling rule interactions in fuzzy production rule reasoning.
• To improve the accuracy and interpretability of fuzzy reasoning systems.

Main Methods:

• Replaced traditional weighted averages with nonadditive nonnegative set functions for rules sharing a consequent.
• Utilized an integral with respect to the nonadditive set function for drawing reasoning conclusions.
• Investigated data-driven methods for determining the nonadditive set function when expert specification is not feasible.

Main Results:

• The proposed integral-based approach effectively models and handles interactions among fuzzy production rules.
• This method leads to a better understanding of the rule base compared to weighted average models.
• Improved reasoning accuracy was observed by addressing the inherent rule interactions.

Conclusions:

• The nonadditive set function and integral method offer a more robust framework for fuzzy production rule reasoning.
• This approach enhances both the accuracy and interpretability of fuzzy inference systems.
• The study provides a pathway for data-driven determination of crucial parameters in fuzzy systems.