Forecasting visitor volume with spatiotemporal models of big data

Abstract

Accurate visitor volume forecasting is essential for the effective management of tourism attractions, particularly during peak periods. Although previous studies have recognized the importance of incorporating spatial relationships into forecasting processes, the spatiotemporal associations between big data–based exogenous variables and tourism demand remain unexplored. This study tests six spatiotemporal models that integrate the spatiotemporal connections of big data variables. Two distinct forecasting exercises—attractions in Beijing, China (pre-COVID-19) and hotels in Orlando, United States (post-COVID-19)—are employed to assess the efficacy of the proposed models. The results demonstrate that incorporating the spatiotemporal interaction of big data improves forecasting performance, suggesting that big data variables can be included in tourism forecasting methods spatially to increase forecasting accuracy. Additionally, the proposed model produces more accurate forecasts for attractions that primarily serve local residents, compared to those that cater predominantly to tourists.