General-to-specific modelling with Bayesian bootstrap aggregation in tourism demand
This study proposes a new forecasting method with a view to further improving the forecasting performance of the existing models built by the investigators. The new method integrates a statistical technique called Bayesian bootstrap aggregation, or BBagging, into the ADL-GETS model selection process. BBagging is designed to reduce forecasting errors through selecting predictors when the decision rules are unstable and the sample size is small. Both empirical and theoretical evidence show that BBagging can push an unstable procedure towards the goal of optimality.
This research represents the first attempt to introduce the BBagging technique into the tourism forecasting field and integrate it with the ADL-GETS model. This research not only has scientific merits, but also significant socio-economic impacts.
Bayesian bagging will be incorporated to the general-to-specific modelling approach to improve the forecasting accuracy.
Professor Haiyan Song
The Hong Kong Polytechnic University