Fonte: Journal of Tourism & Development; v. 1, n. 13/14 (2010); 435-445
Palavras-chave:
Resumo: The modulation and forecast of economic time series related with tourism showed an increase interesting, in last years, due to the relevance of the tourism sector for the Portuguese economy. Hence, the central aim of the present paper consists in the comparative study between the linear regression based model and the Artificial Neural Network (ANN) based model. The inclusion of these two different models has the purpose of understand their potentiality to deal with the peculiar characteristics of the tourism time series such as seasonality and trend. The monthly series that measure the tourism demand “Monthly Guest Nights in Hotels” between January 1990 and December 2008 was used. The developed models achieved a high level of statistical quality of adjustment, and therefore they were used for forecast purposes. A comparison between forecast values and original data for the years of 2007 and 2008 were made. The error, measured by the average of the percentage absolute error (EPAM), for the forecast in that period was 4.2% and 4.1% for the linear regression model and ANN model respectively.
