Genetic design of fuzzy knowledge bases - a study of different approaches

The objective of this work is to design, implement and test two different genetic fuzzy systems approaches with the purpose of analyzing the performance of both when applied to classification problems. In the first approach the fuzzy sets are defined previously by fuzzy clustering and the rule base is automatically generated and optimized using genetic algorithms. In the second approach the data base is the object of genetic algorithm learning, instead of the rule base. In this case, the rule base is generated by means of an auxiliary method (Wang & Mendell). Investigations of both methods developed earlier by the authors are described and then, the results of the comparison experiments performed in the present work are presented. The methods have been selected for investigation with the objective of analyzing the performance and the size of the resulting knowledge bases generated through genetic algorithms applied to different KB components.