MODELLING TOURISM DEMAND:
A COMPARATIVE STUDY BETWEEN ARTIFICIAL NEURAL NETWORKS AND THE BOX-JENKINS METHODOLOGY

by Fernandez, Paula; Teixeira, Joao; Ferreira, Joao and Azevedo, Susana G.
Published in Romanian Journal of Economic Forecasting
, 2008, volume 9 issue 3, 30-50.

 Requires a PDF viewer such as Xpdf or Adobe Acrobat Reader
 173
Kb

Abstract

This study seeks to investigate and highlight the usefulness of the Artificial Neural Networks (ANN) methodology as an alternative to the Box-Jenkins methodology in analysing tourism demand. To this end, each of the above-mentioned methodologies is centred on the treatment, analysis and modelling of the tourism time series: “Nights Spent in Hotel Accommodation per Month”, recorded in the period from January 1987 to December 2006, since this is one of the variables that best expresses effective demand. The study was undertaken for the North and Centre regions of Portugal. The results showed that the model produced by using the ANN methodology presented satisfactory statistical and adjustment qualities, suggesting that it is suitable for modelling and forecasting the reference series, when compared with the model produced by using the Box-Jenkins methodology. 

Keywords: Artificial Neural Networks; ARIMA Models; Time Series Forecasting
JEL Classification: C01, C02, C22, C45, L83