Skip to Main Content
A multiobjective genetic fuzzy system for classification problems is presented. Its advantage is that it uses 3-parameter membership function tuning with dynamic constraints. Therefore, the accuracy is improved without deteriorating the transparency of fuzzy partitions. The initial population is created with a method, which reduces the search space by removing irrelevant input variables. Then, multiobjective genetic fuzzy system optimizes the accuracy and complexity of the fuzzy classifiers and results into Pareto optimal set of compact and accurate fuzzy classifiers, which have transparent fuzzy partitions. The approach is compared to another multiobjective genetic fuzzy system and the advantages of our approach are shown.