40 - Techniques and Applications of Genetic Algorithm-Based Methods for Designing Compact Fuzzy Classification Systems

Abstract

This chapter presents a discussion on the techniques and applications of genetic algorithm-based methods for designing compact fuzzy classification systems. The chapter discusses three approaches to the design of compact fuzzy rule-based systems for multidimensional pattern classification problems. One approach is a rule selection method in which a small number of fuzzy if-then rules are selected by genetic algorithms from a large number of candidate rules. Another approach is a fuzzy classifier system in which a population of fuzzy if-then rules is evolved by genetic algorithms. The other approach is a genetics-based fuzzy machine learning algorithm in which a population of fuzzy rule-based classification systems is evolved by genetic algorithms. The chapter evaluates the performance of each algorithm by computer simulations on the wine data and the credit approval data. The chapter presents several computer simulations in which the fuzzy classifier system showed the best performance among the three algorithms, with respect to classification ability as well as computation time. The genetics-based fuzzy machine-learning algorithm based on the Pittsburgh approach could not find good rule sets. Simulation results suggested that the rule selection methods are useful as a knowledge acquisition tool because the number of selected fuzzy if-then rules was very small. The simulation results by the rule selection method on the credit approval data were not as good as those by the fuzzy classifier system.