2000 Volume 12 Issue 6 Pages 806-815
Fuzzy modeling is a method to describe input-output relationships using fuzzy inference. Fuzzy modeling can approximate complex nonlinear systems and it is easy for humans to understand fuzzy rules. One of the major difficulties in fuzzy modeling is allocation of membership functions for each input variable. This paper proposes a new method for allocation of membership functions. For proposing the method in this paper, case studies on methods for initial allocation of membership functions, ones for insertion of membership functions, and ones for stopping conditions, are carried out. The initial allocation of membership functions is searched for by roughly allocating them under Akaike Information Crirerion(AIC). The stopping condition of the learning of fuzzy models is given either by a criterion using test data with sample data being divided into two groups:training data and test data, or by AIC where all the samples are used as training data. Experiments are done to show that the proposed method is effective in allocating membership functions more properly than the conventional method.
