A Survey on the Design of Fuzzy Classifiers Using Multi-Objective Evolutionary Algorithms

Fuzzy systems have been used in many fields like data mining, regression, patter recognition, classification and control due to their property of handling uncertainty and explaining the property of complex system without involving a specific mathematical model. Fuzzy rule based systems (FRBS) or fuzzy rule based classifiers (particularly designed for classification purpose) are basically the fuzzy systems which consist a set of logical fuzzy rules and These FRBS are annex of traditional rule based systems, because they deal with “If-then” rules. During the design of any fuzzy systems, there are two main features, Interpretability and accuracy which are conflicting with each other i.e. Improvement in any of these two features causes the decrement in another one. This condition is called Interpretability –Accuracy Trade-off. To handle such kind of situation Multi Objective Evolutionary Algorithms are used to design fuzzy systems. This paper is a review of different design approaches of fuzzy systems and various methods to analyze the I-A tradeoffs in the design of fuzzy classifiers. Also various techniques for assessment of accuracy and interpretability have been discussed.