Defuzzification: criteria and classification (Q1965292)

The paper is devoted to the evaluation of various defuzzification methods encountered in the existing literature. By defuzzification we mean a transformation of a fuzzy set into its numeric representative. General evaluation criteria are outlined (including core selection, scale invariance, monotony, triangular conorm criterion, continuity as well as computational complexity). The taxonomy of the defuzzification methods under discussion includes three main groups: maxima methods, distribution methods, and area methods. It has been found that the maxima methods conform to the basic criteria such as core selection, scale invariance, monotony, and a triangular conorm criterion. The two other groups do not satisfy these criteria but provide the continuity property.