Fuzzification and Defuzzification Process in Genetically Evolved Fuzzy Cognitive Maps (GEFCMs)

This paper describes the fuzzification and defuzzification process in the framework of hybrid systems comprising Fuzzy Cognitive Maps (FCMs) and Genetic Algorithms (GAs). More specifically, it provides a stepwise methodology for fuzzification and defuzzification aiming at both an improved approach of the human reasoning pattern and an increase of the decision-making potentials. The fuzzification process is primarily based on producing fuzzy information provided by a group of experts. Each concept is analyzed into trapezoidal membership functions of either fixed or variable widths, with these intervals labeled and stored for the defuzzification process later on, during which the levels are matched according to the membership functions of each concept. The defuzzification process is more complicated than the fuzzification one and consists of four basic iterative stages: The Iteration, the Max-Min Average Computation, the Categorization and, finally, the Realization Inference Stage.