Particle swarm optimization and least squares estimaton of NARMAX

• Published in: ARPN Journal of Engineering and Applied Sciences
• Main Author: Abdullah S.M.; Yassin A.I.M.; Tahir N.M.
• Format: Article
• Language: English
• Published: Asian Research Publishing Network 2015
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-84953401910&partnerID=40&md5=91c5952268558e12c0f1f2d543a55784

SI process consist of three steps; structure selection, parameter estimation, model validation. This paper compared method of Particle Swarm Optimization (PSO) and Linear Least Squares solution methods (LLS) (Normal Equation (NE), QR decomposition (QR) and Singular Value Decomposition (SVD)) for parameter estimation using polynomial NARMAX models. The comparison was tested on Flexible Robot Arm (FRA) dataset. Our analysis suggests that the PSO algorithm is comparable to other established algorithms for LLS parameter estimation in terms of model fit accuracy and information criteria (Akaike Information Criterion (AIC), Final Prediction Error (FPE) and Model Descriptor Length (MDL)). Additionally, the PSO algorithm was found to slightly improve the correlation tests relative to other LLS tested algorithms. © 2006-2015 Asian Research Publishing Network (ARPN).