FORECASTING INTERREGIONAL COMMODITY FLOWS USING ARTIFICIAL NEURAL NETWORKS: AN EVALUATION

The aim of this paper is to evaluate empirically the predictive performance of artificial neural network (ANN) models in the area of commodity flows with respect to an existing benchmark model. While earlier studies in this area of research focused on the training and validation of ANNs, this study focuses on the predictive performance of ANNs for different commodity groups. The paper first presents an outline of commodity flows as a form of spatial interaction. A basic discussion about ANNs is included, and the data and variable sets used in the study are presented. The clearest finding of the study is that while a spatial interaction ANN model proves astonishing calibration superiority with respect to a conventional regression based Box-Cox model, the predictive performance of the same model outperforms that of the ANN nodel for interregional commodity flows.

• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/oclc/1767712
• Corporate Authors:

  Taylor & Francis

  4 Park Square, Milton Park
  Abingdon,   United Kingdom  OX14 4RN
• Authors:
  • Celik, H M
• Publication Date: 2004-12

Language

• English

Media Info

• Features: Appendices; Figures; References; Tables;
• Pagination: p. 449-467
• Serial:
  • Transportation Planning and Technology
  • Volume: 27
  • Issue Number: 6
  • Publisher: Taylor & Francis
  • ISSN: 0308-1060
  • Serial URL: http://www.tandf.co.uk/journals/titles/03081060.html

Subject/Index Terms

• TRT Terms: Commodity flow; Forecasting; Freight transportation; Neural networks; Regional transportation
• Subject Areas: Freight Transportation; Highways; Planning and Forecasting;

Filing Info

• Accession Number: 00986711
• Record Type: Publication
• Files: TRIS
• Created Date: Feb 4 2005 12:00AM