A Heuristic Approach of Model Selection in Multiple Nonlinear Regression Analysis

Proceedings of the IADIS International Conference "Applied Computing 2006" 2006
Gints Jēkabsons, Jurijs Lavendels

This paper reflects a research goal of which is to develop heuristic approach for multiple nonlinear regression analysis model selection. From sixteen heuristic search algorithms suitable for multiple nonlinear regression analysis eight most popular algorithms were considered. All of the algorithms were classified and empirically evaluated from the aspect of both necessary computing resources and optimality of the results. The theoretical results of the research are implemented in software, which was used for approbation of the described approach in construction behavior modeling applications at Institute of Materials and Structures, Riga Technical University. Models obtained were more effective than previously used. Developed software is effective and competitive tool for solving practical regression problems.

Atslēgas vārdi
Regression, approximation, model selection, heuristic search methods
Hipersaite
http://www.iadis.net/dl/final_uploads/200601C026.pdf

Jēkabsons, G., Lavendels, J. A Heuristic Approach of Model Selection in Multiple Nonlinear Regression Analysis. No: Proceedings of the IADIS International Conference "Applied Computing 2006", Spānija, San Sebastian, 4.-6. marts, 2006. San Sebastian: IADIS Press, 2006, 524.-527.lpp. ISBN 972-8924-09-7.

Publikācijas valoda
English (en)