TSK Fuzzy Modeling Based on Kernelized Fuzzy Clustering and Least Squares Support Vector Machines

In this paper, a novel learning method based on kernelized fuzzy clustering and least squares support vector machines (LSSVM) is presented to improve the generalization ability of a Takagi-Sugeno-Kang (TSK) fuzzy modeling. Firstly, the fuzzy partition of the product space of input and output is obtained by kernelized fuzzy clustering. Then, a computationally efficient numerical method is proposed. In the proposed algorithm, the fuzzy kernel is generated by premise membership functions. Numerical experiments show that the presented algorithm improves the generalization ability and robustness of TSK fuzzy models compared with traditional learning methods and LSSVM.