Extraction of Fuzzy Knowledge Bases from Experimental Data by Genetic Algorithms

This paper deals with the problem of determination of linguistic "IF-THEN" rules from available experimental data, which is inverse to the problem of identification of nonlinear dependences by fuzzy knowledge bases. A method of genetic algorithms is proposed. The method is based on the operations of crossover, mutation, and selection of initial variants of solutions or so-called chromosomes, from which the most optimal solutions are subsequently chosen. The method is illustrated by a computer experiment consisting of the determination of knowledge on a nonlinear object with two input variables and one output variable.