Improved Weighted Fuzzy Reasoning Algorithm Based on Particle Swarm Optimization

This paper proposes an improved weighted fuzzy reasoning algorithm based on particle swarm optimization (PSO) for handling classification problems. Fuzzy production rules of rule-based system are used for knowledge representation, where the local and global weights appearing in the rules are represented by real values between zero and one. In order to model the overlapping existing among the rules sets corresponding to different classes, this paper proposes a new set function to draw the reasoning conclusion, with respect to a non-additive nonnegative set function and the weights of the rules determined by PSO. And the criterion of the parameters adjustment is based on maximum fuzzy entropy principle, which can overcome the shortcoming of over-fitting. An experimental investigation is performed on the UCI datasets and the encouraging result shows that the proposed algorithm based on PSO can strengthen the reasoning capability of rule-based system.