Finding useful fuzzy concepts for pattern classification using genetic algorithm

In this paper, a fuzzy classifier is treated as a fuzzy information retrieval system. To construct such a system, a fuzzy data mining method is employed to determine useful fuzzy concepts. Subsequently, each of the classes and patterns can be represented by a fuzzy set of useful fuzzy concepts. From the viewpoint of fuzzy information retrieval, a pattern can be categorized into one class if there exists a maximum degree of similarity between them. The genetic algorithm (GA), whose objective is to find a compact set consisting of useful fuzzy concepts with high classification capability, is further employed to automatically determine parameter specifications that are not easily specified by users. To evaluate classification performance of the proposed method, computer simulations are performed on some well-known classification problems, demonstrating that the generalization ability of the proposed method is comparable to other fuzzy classification methods.