Article ID: | iaor2008593 |
Country: | United Kingdom |
Volume: | 24 |
Issue: | 7 |
Start Page Number: | 505 |
End Page Number: | 521 |
Publication Date: | Nov 2005 |
Journal: | International Journal of Forecasting |
Authors: | Bunn Derek W., Pantelidaki Stefania |
Keywords: | retailing, neural networks |
This paper proposes an approach that models and forecasts sales through a flexible parametric response function (multifunctional), allowing for differentiated behavioural assumptions of the response determinants to be specified, and uses neural network modelling as a re-specification tool for the response model in order to improve forecasting performance. An initial experiment on a sample of sales data demonstrates feasibility and gives comparative insights via alternative model specifications.