Tourism Demand Probabilistic Interval Forecasting Approach With Integrated Deep PatchTST‐GMM Model
Zhongshan Cui, Hui Li, Shaolong Sun, Shouyang WangABSTRACT
Tourism demand forecasting is an important guide for decision‐making in related industries, especially because interval forecasting can provide managers with more decision‐relevant information than a single point estimate. In order to better forecast tourism demand, this study proposes a probabilistic interval forecasting framework for tourist demand by considering sequence instability and possible distributional drift characteristics. Specifically, the framework uses PatchTST to capture local temporal dependencies and enhance robustness under nonstationary patterns and then uses a Gaussian mixture model (GMM) to estimate the predictive distribution from which decision‐oriented prediction intervals are derived. The results show that the interval prediction results are improved over the benchmark models under the same time series input setting, and the findings can provide richer decision support for relevant managers.