An Improvement in Fuzzy Inference System for Classification Problems Using Mutual Information

Fuzzy Inference System (FIS), is one of the most powerful inference systems that is widely used in field of classification. Indeed, in this approach, FIS is engaged to create a mapping from features (inputs) to classes (outputs) using fuzzy set theory. So far, many efforts have been gone into improving classification accuracy that has been performed by FIS. Generally, these efforts have been conducted in the following areas: efficient fuzzy rule generation, fuzzy membership function tuning, fuzzy rule weight tuning, feature selection for antecedent part of fuzzy rules, and so on. In this paper we consider this issue and propose a method based on mutual information for applying impact factor of input parameters on fuzzy inference process for improving accuracy of fuzzy classification. Finally, we test our proposed method for boosting classification on six different problems using manual and auto generated FIS. Classification results confirm correctness of the proposed method and are so promising.