An algorithm for optimizing membership functions of fuzzy inference systems based on fuzzy associative memory

Abstract

A fuzzy associative memory is a fuzzy logic tool for pattern recognition or control problems. Fuzzy inference systems based on fuzzy associative memory have a wide range of practical applications. These systems need to be determined how to form and how many membership functions for each input variable by analyzing its histogram. This paper proposes an algorithm for optimizing their membership functions by a threshold set and a method that finds out optimal membership functions whereby classification effects of the fuzzy inference systems can be improved. The algorithm is associated with a measure of useful degree of input membership functions to increase accurate classification rates of the system. The paper also based on the experiment to show criteria for collecting training data to improve the effect of recognition or classification of the fuzzy inference systems. To confirm the effectiveness, the proposed algorithm is applied to a pattern recognition problem with the iris data through a fuzzy inference system based on fuzzy associative memory.