Article ID: | iaor1993317 |
Country: | United Kingdom |
Volume: | 19 |
Start Page Number: | 179 |
End Page Number: | 189 |
Publication Date: | Sep 1992 |
Journal: | Computers and Operations Research |
Authors: | Ignizio James P., Burke Laura I. |
Keywords: | philosophy, artificial intelligence |
Problems that seem to be encountered, time and time again, by the practicing Operational Research OR analyst include: (i) resource allocation (i.e. continuous and discrete optimization); (ii) classification (i.e. pattern recognition, or discriminant analysis); (iii) prediction/estimation; and (iv) clustering. When faced with such problems, the Operational Research OR analyst has a number of conventional methods to choose from, such as: linear programming, discrete optimization, statistically based discriminant analysis, regression and cluster analysis. In this paper the authors will address yet another alternative: the employment of neural networks. While certainly no panacea, the neural network approach may, in certain instances, offer advantages that range from minor to substantial. Further, a neural network approach permits solution by means of parallel processing, thus providing certain unique and significant advantages that are inherent to distributed computing. As such, the Operational Research OR analyst who remains unfamiliar with this approach cannot, the authors believe, consider himself or herself to be fully prepared for the most effective solution of a variety of problems, both now and in the future. In this paper, the authors shall introduce the neural networks approach, from an Operational Research OR perspective-and indicate just where and how such a tool might find application.