ABSTRACT
This chapter proposes a genetic-algorithm-based method for constructing a fuzzy classification system with linguistic if-then rules. In this method, first a large number of linguistic rules are generated from numerical data, that is, from training patterns. Then only a small number of significant rules are selected by a genetic algorithm in order to construct a compact fuzzy classification system. A set of linguistic rules is coded as an individual in the genetic algorithm. The proposed method is illustrated by computer simulations on a numerical example and the well-known iris data. The chapter also proposes a hybrid approach that incorporates a learning procedure into the genetic algorithm. The grade of certainty of each linguistic rule in each individual is adjusted by the learning procedure during the execution of the genetic algorithm. The chapter demonstrates by computer simulations on the iris data that the hybrid algorithm can find a small number of linguistic rules with high classification power.
