Genetic Rule Selection as a Postprocessing Procedure in Fuzzy Data Mining

Abstract

We examine the effect of genetic rule selection as a postprocessing procedure in fuzzy data mining. Usually a large number of fuzzy rules are extracted in a heuristic manner from numerical data using a rule evaluation criterion in fuzzy data mining. It is, however, very difficult for human users to understand thousands of fuzzy rules. Thus it is necessary to decrease the number of extracted fuzzy rules when our task is to present understandable knowledge to human users. In this paper, we use genetic rule selection to decrease the number of extracted fuzzy rules. Through computational experiments, we examine the effect of genetic rule selection. First we extract fuzzy rules that satisfy minimum support and confidence levels. Thousands of fuzzy rules are extracted from numerical data in a heuristic manner. Then we apply genetic rule selection to extracted fuzzy rules. Experimental results show that genetic rule selection significantly decreases the number of extracted fuzzy rules without degrading their classification accuracy