Regional tourism demand forecasting with spatiotemporal interactions: a multivariate decomposition deep learning model

ABSTRACT

With the advancement of economic globalization and regional integration, regional tourism flows are more closely linked, which provides new clues for improving forecasting. This study develops a multivariate decomposition deep learning model to forecast tourism demand by capturing spatiotemporal interactions among regional tourism flows. The multivariate decomposition technique is introduced to reduce data complexity, while convolutional neural networks and long short-term memory networks are extracting spatial and temporal correlations of regional tourism flows. The effectiveness of the model is demonstrated in two heterogeneous international tourism cases of tourist arrivals from China or Japan to leading destinations in Southeast Asia.

KEYWORDS:

• Regional tourism demand forecasting
• international tourist arrivals
• spatiotemporal interactions
• multivariate decomposition
• deep learning
• southeast Asia

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by China Scholarship Council: [Grant Number 202206280175, 202206280179]; National Key Research and Development Program of China: [Grant Number 2022YFF0903000]; National Natural Science Foundation of China: [Grant Number No. 72101197]; Natural Science Basic Research Program of Shaanxi Province: [Grant Number 2023-JC-QN-0785]; Science and Technology Project of China Huaneng: [Grant Number HNKJ20-H87]; National Natural Science Foundation of China: [Grant Number No. 71774130].