Fuzzy Rule Base Simplification via Conflict Resolution by Aggregating Rule Consequents

Fuzzy systems have been widely applied to many real-world applications. Whilst successful, these applications have revealed certain significant limitations of such systems. Amongst the questions raised is how to handle rule base complexity, especially when dealing with sophisticated domain problems. This paper proposes a method to simplify fuzzy rule bases by reducing the occurrence of inconsistent rules (which have the same antecedents but different consequences). In particular, it implements the rule base simplification task by handling conflict rules, via aggregating different consequent fuzzy sets of inconsistent rules with linear combination. This learns from the underlying ideas of fuzzy rule interpolation, which is simple and easy to understand while being effective. Experimental studies show that the proposed method can not only resolve inconsistencies embedded in a fuzzy rule base but also empower the fuzzy rule base to achieve better performance for rule bases learned from data.