A comparison of linear and non linear models to forecast the tourism demand in the North of Portugal

Abstract

Abstract In order to contribute for enriching studies in the tourism field, it was intended with this research paper performing the comparison between the model based on linear regression and the model based on artificial neural networks and analyses of the performance of those models. Additionally, the usefulness of the time series that measures the number of hours of Sunshine should be confirmed. We used for this purpose the monthly series that measures the demand for tourism: “Monthly Nights in Hotels in the Northern Region of Portugal”, recorded in the period from January 1990 to December 2009. A linear regression model based on the first differences was developed producing none statistical infractions. A previously developed ANN based model was applied for the new period of time under comparison. Both models have the sunshine time series in their entrance. Both methodologies proved to achieve similarly good results in getting the seasonality of the time series, because the correlation coefficient was at the level of 0.99. Also both models could predict with high quality the magnitude of the time series because the mean absolute percentage error was 4.1% and 3.5% for the linear model and for the ANN based model, respectively.

Keywords

• Forecasting
• Tourism Demand
• General Linear Model
• Artificial Neural Networks