Technique for designing accurate and compact Takagi–Sugeno fuzzy systems

Authors: Hodashinskiy I. A., Carin K. S.

Annotation: The paper presents the new technique to design accurate and compact Takagi–Sugeno fuzzy systems. The technique includes the following steps: forming the initial data, the generation of fuzzy system structure, optimization of parameters consequent and antecedents of fuzzy rules. The compromise between compactness and accuracy of the model is achieved by using three statistical information criteria: Akaike, Bayesian, Hannan–Quinn Information Criterion. Results of experiments on the approximation of the actual data are delivered. Simulation experiments show validation of the proposed technique.

Keywords: takagi–sugeno fuzzy systems, dynamic partitioning, piecewise linear initialization, cuckoo search algorithm, recursive least squares method, statistical information criteria