INITIALIZATION METHOD FOR FUZZY TAKAGI-SUGENO SYSTEMS
I. A. Hodashinsky, K. S. Sarin, S. A. Cherepanov
Tomsk State University of Control Systems and Radio Electronics, pr. Lenina 40, Tomsk, 634050 Russia
Keywords: инициализация, нечёткие системы, динамическое разбиение, рекуррентный метод наименьших квадратов, initialization, fuzzy systems, dynamic partitioning, recursive least squares method
Abstract
This paper presents a method for initializing fuzzy Takagi-Sugeno systems in which the initial values of fuzzy antecedents are obtained by dynamic decomposition of the input space, and consequent values by the recursive least squares method. The results of experiments on 13 datasets from the KEEL repository are described. Comparisons with the results of their approximation by five known identification algorithm are given.
|
