A Random Forest based Learning Framework for Tourism Demand Forecasting with Search Queries

This study proposes a novel framework for tourism demand forecasting, which combines search queries generated on the Internet, advanced feature selection methods, and machine learning based forecasting technique. This new methodology is applied to forecast tourism demand in two popular destinations in China: Beijing and Haikou. The study evaluates the performances of various feature selection approaches in tourism forecasting. We further show that the random forest feature selection and support vector regression with radial basis function can be extremely useful for the accurate forecasts of tourist volumes. This research highlights the advantage of big data and machine learning algorithms in the tourism forecasting.