A fuzzy clustering based approach for generating interpretable fuzzy models

A systematic fuzzy modeling approach considering both accuracy and interpretability is developed in this paper. First, fuzzy clustering algorithm, combined with least square method, is used to obtain initial fuzzy model with excessive rules. Then rule reduction is performed by orthogonal least squares to obtain simplified fuzzy model with high interpretability. Finally, genetic algorithm is adopted to optimize the simplified fuzzy model to improve accuracy. The proposed approach is successfully applied to a real world coke-oven temperature system, and the result shows its effectiveness.