Fuzzy Rules Generation using Conditional Clustering

Abstract

This work describes the Fuzzy-CCM (Fuzzy Conditional Clustering based Modeling) method, which generates fuzzy rules automatically using the conditional Fuzzy C-Means algorithm, and proposes the use of a new strategy for attributes combination on the context definition step, using heuristic search based on the criteria of best performance. This strategy allows the reduction of the number of contexts considered and avoids the generation of rule bases that do not present performance improvements with relation to the already known ones. The main goal of the Fuzzy-CCM method is to provide new manners to deal with the issue of interpretability of rule bases. The generated rules have a format different from the usual that combines linguistic variables and groups in the antecedent. The experiments were performed using data sets with discrete and continuous output values, in order to evaluate the proposed approach and to compare the results with those obtained with the Wang-Mendel and Fuzzy C-Means methods. The advantages of the method, the benefits of the proposed strategy and the results obtained in the experiments are discussed