A Spatial-temporal Model for Tourism Demand Forecasting.

Accurate forecasting of tourism demand is important to the development of tourism. However, the difficulties in recognizing complex spatial and temporal features make it challenging to accurately forecast tourism demand. In addition, existing methods are not practical and flexible enough since they usually established multiple models for different scenic spots. In this paper, we propose a novel method for tourism demand forecasting based on the fully connected long short-term neural network, which enables simultaneous identification of spatial and temporal features for better forecasting accuracy. To enhance the practicality and flexibility of our method, we propose to establish one general model for multiple scenic spots. Experimental results demonstrate that the proposed method outperforms other models in the daily tourism demand forecasting for the Wanshan Archipelago, an emerging tourism spot in Zhuhai, China.