Fuzzy rule modeling based on FCM and support vector regression

To design a fuzzy rule-based modeling framework with good generalization ability has been an active research topic for a long time. As a powerful machine learning approach for function approximation and regression estimation problems, support vector regression (SVR) is known to have good generalization ability. In this paper, we adopt the FCM clustering algorithm to group data patterns into clusters, after FCM clustering, the membership grade are applied to generate fuzzy kernel. Then, the support vector learning with fuzzy kernel provides a fuzzy IF-THEN rules architecture. In terms of fuzzy rules, the overall fuzzy inference system can be calculated by weighting the inferred output values from each cluster with their corresponding membership values. Experimental results show that the proposed method can achieve good approximation performance.