Title: Modelling empirical data and decision making with neural networks

Authors: Zoran Vojinovic, Vojislav Kecman, Rainer Seidel

Addresses: University of Auckland, School of Engineering, Private Bag 92019, Auckland, New Zealand. University of Auckland, School of Engineering, Private Bag 92019, Auckland, New Zealand. University of Auckland, School of Engineering, Private Bag 92019, Auckland, New Zealand

Abstract: This paper discusses neural networks as a replacement for traditional statistical forecasting and regression based decision models. From the vast literature and studies on neural networks, one may find that some authors advocate neural networks as a promising tool, whilst other authors are concerned that neural networks might be oversold and are not certain under what conditions they are better than traditional methods. Our intention here is to provide an overview on the difference between linear and non-linear modelling approaches (such as neural networks) and to provide a review of the literature with directions for future research. In doing this, we have summarised our findings from the literature and from several studies that we have performed. We found that the majority of empirical studies to date show that neural networks perform better than traditional linear methods and therefore have great potential to replace traditional forecasting and decision-making models. However, more research is needed to enable neural networks to become a standard tool for applications across different fields.

Keywords: Neural networks; linear models; forecasting; regression based decision models.

DOI: 10.1504/IJMDM.2002.002472

International Journal of Management and Decision Making, 2002 Vol.3 No.2, pp.180-202

Published online: 18 Jul 2003 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article