Approximate pattern classification with fuzzy boundary

A neural-network-based approximate classification method with a fuzzy boundary is proposed and the advantage of the proposed method is demonstrated by the application to the iris data of Fisher. Conventional classification problems can be described as finding a clear cut-off boundary to divide the pattern space into disjoint decision areas. In a real problem, it is not always easy, if not impossible, to find a sharp boundary between decision areas. Therefore we propose an approximate classification method where the existence of a boundary area (i.e., fuzzy boundary) between the decision areas is assumed.