Correlation and Prediction for the Portuguese Tourism Sector Leveraged by Google Trends Data

In the last decade, tourists gradually became more connected, gathering information about their destinations on the Internet. This suggests that the readily available search data is increasingly more relevant for up to date descriptive and predictive analysis regarding tourism data. This benefits the tourism sector, which is a very important sector for the economy of many countries, including Portugal. In this work, the relation between guests’ data in tourism accommodation in Portugal and the search data of the main destinations in the country is analyzed for the last few years. The search data used is from Google Trends, considering different lag values. Short horizon tourist demand forecasts were produced considering exponential smoothing (ETS), autoregressive integrated moving average (ARIMA) and dynamic regression models, through a cross-validation experimental setup. It was evaluated if the inclusion of search data improved the forecasts. The study was performed considering the eight main guests’ countries of origin, which are: Portugal, Great Britain, Ireland, Spain, France, Germany, Brazil, and Italy. The results indicate correlation between moderate and very strong for all countries. Furthermore, it was possible to observe that the correlation was higher for almost all the countries when the search data considered was lagged between 0 and 3 months. Regarding the comparative analysis between the forecasting models, the ARIMA models presented the best results for most of the countries.