Bagging in Tourism Demand Modeling and Forecasting

Abstract

This study introduces bootstrap aggregation (bagging) in modeling and forecasting tourism demand. The aim is to improve the forecast accuracy of predictive regressions while considering fully automated variable selection processes which are particularly useful in industry applications. The procedures considered for variable selection is the general-to-specific (GETS) approach based on statistical inference and stepwise search procedures based on a measure of predictive accuracy (MPA). The evidence based on tourist arrivals from six source markets to Australia overwhelmingly suggests that bagging is effective for improving the forecasting accuracy of the models considered.