O. Kaynak et al. (Eds.): ICANN/ICONIP 2003, LNCS 2714, pp. 225–233, 2003. © Springer-Verlag Berlin Heidelberg 2003 Fuzzy Model Identification Using Support Vector Clustering Method $\úHJOUçar 1 , Yakup Demir 1 , and Cüneyt *]HOLú 2 1 Electrical and Electronics Engineering Department, Engineering Faculty, )ÕUDW University, Elazig, Turkey agulucar@ieee.org,ydemir@firat.edu.tr 2 Electrical and Electronics Engineering Department, Dokuz Eylül University, Kaynaklar Campus, ø]PLU7XUNH\ guzelis@eee.deu.edu.tr Abstract. We have observed that the support vector clustering method proposed by Asa Ben Hur, David Horn, Hava T. Siegelmann, Vladimir Vapnik, (Journal of Machine Learning Research, (2001), 125–137) can provide cluster bounda- ries of arbitrary shape based on a Gaussian kernel abstaining from explicit cal- culations in the high-dimensional feature space. This allows us to apply the method to the training set for building a fuzzy model. In this paper, we sug- gested a novel method for fuzzy model identification. The premise parameters of rules of the model are identified by the support vector clustering method while the consequent ones are tuned by the least squares method. Our model does not employ any additional method for parameter optimization after the ini- tial model parameters are generated. It gives also promising performances in terms of a large number of rules. We compared the effectiveness and efficiency of our model to the fuzzy neural networks generated by various input space- partition techniques and some other networks. 1 Introduction In recent years, fuzzy models have successfully appeared on a lot of applications in system identification, control, prediction and inference. An important property of fuzzy models is their ability to represent highly nonlinear systems. In comparison with other nonlinear black-box modeling techniques, fuzzy models have the advantage of giving insight into the relations between model variables and combining prior knowl- edge with the information identified from numerical data. The fuzzy models are accomplished by structure identification and parameter ad- justment procedures. Generally, these models are built from two learning phases, the structure learning phase and the parameter learning phase. These two phases are usu- ally done sequentially; the structure learning phase is employed to decide the structure of fuzzy rules first and then the parameter learning phase is used to fine tune the coef- ficients of each rule obtained from the first one. Various methods have been proposed to solve these problems separately or in a combinatorial way [1–3].