Research on clustering algorithm of fuzzy structure identification

The adaptive neuro-fuzzy inference system (ANFIS) has good advantages over conventional modeling systems for nonlinear processes. It constructs fuzzy language models to approach the features of the process and uses neural network to update the fuzzy parameters. Since it makes a linear division to the input space, the number of the fuzzy rules will multiple when meeting complex processes. Also. ANFIS is easy to trap into local minima. To solve these problems, the paper proposed a new kind of clustering algorithm based on artificial immune system (AIS) for modeling of ANFIS. In this algorithm, the immune network does cloning, mutation and suppression actions to antibodies and memory data sets, extracts useful fuzzy rules from them and avoids training possibilities of ANFIS to local minimal ports. The paper deeply discusses the randomness of AIS that affects on clustering stability, and analyses its problems in clustering speed, then makes necessary modifications. Through simulation, good identification results are obtained.