Model evaluation and selection in multiple nonlinear regression analysis

Abstract

The main problem in regression model selection is finding the best model that best fits the data, i.e. it does not neither overfit nor underfit. The aim of this work is to show one of possible ways to find adequate nonlinear regression models (parametric) of technical systems based on an heuristic search and analytical optimality evaluation approach by taking into consideration the computational power of modern computers. © 2007, Taylor & Francis Group, LLC.