Abstract
In this paper, we propose the heuristic clustering for a fuzzy modeling. This approach devide an input variable space into some clusters by using an approach following human's experience and intuition. As input variable space is devided into a lot of subspaces, and we suppose that their relations of input variables and output variables are linear. If unit normal vectors of some subspaces are similar, their are unified as a cluster. Then, some clusters are generated. After the heuristic clustering, we set membership functions of premise parts, and values of consequent parts are learned by a steepest decent method. To verify this approach, we show nonlinear function's approximation as a result. Furthermore, as an example of application, we show fuzzy model for the predicting diagnosis of ectopic pregnancy after the pregnancy effect for infertile patients.
