Title :
Recursive least-squares method with membership functions
Author_Institution :
Inst. of Syst. Eng., Zhejiang Univ., Hangzhou, China
Abstract :
Linear regression with the recursive least-squares algorithm is widely used for parameter estimation. However, in a number of cases, linear regression is not sufficient to accurately correlate a process. A method combining the recursive least-squares algorithm and membership functions is presented to overcome the shortcoming of traditional linear regression in that multiple linear sub-models with different coefficients are used. The model coefficients can be tuned recursively according to the membership functions preset for the models. Simulation results show that this method can be used with good performance for nonlinear process, which is previously difficult to be estimated by using the traditional linear regression.
Keywords :
least squares approximations; nonlinear estimation; recursive estimation; regression analysis; linear regression; membership functions; multiple linear submodel coefficients; nonlinear process; parameter estimation; recursive least squares algorithm; recursive least squares method; Equations; Fuzzy neural networks; Least squares approximation; Linear regression; Parameter estimation; Recursive estimation; Systems engineering and theory; Time varying systems; Training data; Vectors;
Conference_Titel :
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on
Print_ISBN :
0-7803-8403-2
DOI :
10.1109/ICMLC.2004.1382101