Fuzzy Classifier Based On AFS Theory And Genetic Algorithm

With the progress of society, unceasing development of database technology and widely use of database management, data are increased rapidly. While the information people could get is a small segment. Data mining come into being and develop to solve this problem. And classification is a key issue of data mining; Therefore the research on classification is very important.The classifier of this paper is based on AFS theory and Genetic Algorithm. AFS (Axiomatic Fuzzy Set) theory was proposed firstly as a new analytic method of fuzzy mathematics. In the framework of AFS theory, the membership functions and their logic operations for fuzzy concepts can be impersonally determined according to original data and facts.Genetic Algorithm (GA) is based on natural selection and genetic theory. It is an effect global searching algorithm that combines the rule of survival of the fittest and the random information change mechanism in the chromosome of colony.Through the research on AFS theory and Genetic Algorithm, by virtue of the advantage of the AFS theory that the membership function are directly determined by the information of original data and it need not to be adjusted to solve the problem and the global search characteristic of the Genetic Algorithm. The advantages of them are fully utilized and a new classifier is proposed based on them. The fuzzy membership function of fuzzy concepts by applying the AFS theory; With the membership function, candidate rules were obtained that are used to classify the samples; These rules are selected by Genetic Algorithm theory, and these rules are combined by AFS theory and these rules are treated as a fuzzy classifier.In order to show the advantage of the proposed method, some experiments are done on three data sets of UCI (University of California, Irvine) machine learning database, such as the iris data, the wine data, and the breast cancer test data. The experimental results show that the proposed fuzzy classifier based on AFS theory and Genetic Algorithm has few rules, high classification rate, and good interpretability. Moreover, compared to releted references, a further analysis of experiment test was given.