Forecasting Tourist Arrivals via Random Forest and Long Short-term Memory

In recent years, deep learning has been attracting substantial attention due to its outstanding forecasting performance. However, the application of deep learning methods in solving the problem of forecasting tourist arrivals has been few. For the efficient allocation of tourism resources, tourist arrivals must be accurately predicted for government and tourism enterprises. In this study, a new hybrid deep learning approach is developed for tourist arrival forecasting. Random forest is used to reduce the dimensionality of the search query index data for selecting a small subset of informative features that contain the information that is most related to the tourist arrivals. Differential evolution algorithm is designed for choosing the lag lengths of each search query index and historical tourist arrival data for reconstructing the forecasting input. Long short-term memory (LSTM) is used for modeling the nonlinear relationship between tourist arrivals and search query index data. Two comparative examples, namely, Beijing City and Jiuzhaigou Valley, are applied for verification of the forecasting accuracy of the proposed deep learning method. The results indicate that the proposed deep learning method outperforms some time series and machine learning methods.