Tourism Demand Forecasting with Multi-terminal Search Query Data and Deep Learning

This study aims to explore the effectiveness of multiple devices’ search query in tourism demand forecasting. Accordingly, this study collects search data from computer and mobile devices, and proposed a hybrid deep learning model, namely MUL-CNN-LSTM, with improved structure accordingly. In the model, a dual CNN module is employed to extract deep features of search queries from multi-devices. Subsequently, the LSTM is applied to generate the prediction of tourism demand. By taking JiuZhaiGou Valley as a case, the empirical results of three groups of comparisons demonstrate that the proposed deep learning model, along with the multi-terminal search query data, significantly enhances forecasting performance.

Duke Scholars

Author Mustafa Misir DKU Faculty