An Interval Decomposition-Ensemble Model for Tourism Forecasting

In order to accurately capture the variability of tourism demand, this paper proposed a new decomposition-ensemble framework for forecasting interval-valued time series (ITS) of tourism arrivals. The procedure consists of four main steps: ITS decomposition, determination of the optimal decomposition technique, component ITS forecasting, and ensemble. The investigation revealed the optimal theoretical approach for choosing the decomposition technique in terms of multi-scale complexity. In addition, a comparison was made between the performance of two types of models that predict the upper and lower limits of ITS separately versus simultaneously. Using the weekly ITSs of tourist arrivals to Mount Siguniang, in western China, and Hawaii, USA, during both COVID and non-COVID periods, an empirical study was conducted to illustrate the framework. The results demonstrated that the proposed model exhibits higher predictive accuracy and greater robustness, compared to other models. This indicates the model's effectiveness in forecasting the ITS of tourism demand.

Automated summary

This paper proposes a new decomposition-ensemble framework for accurately forecasting the variability in tourism arrivals. The framework includes four main steps: ITS decomposition, determination of the optimal decomposition technique, component ITS forecasting, and ensemble. An empirical study comparing different models demonstrates that the proposed model shows higher predictive accuracy and robustness, highlighting its effectiveness in forecasting tourism demand.