Leveraging Social Media Data for Enhanced Forecasting of International Student Arrivals in Australia

This study examines the extent to which incorporating social media data enhances the predictive accuracy of models forecasting international students’ arrivals. Private social media data collected from a public university, along with collected web traffic data and Google Trend data, were used in the forecasting models. Initially, a correlation analysis was conducted, revealing a strong relationship between the institution’s international student enrolment and the social media activity, as well as with the overall number of international students arriving in Australia. Building on these insights, features were derived from the collected data for use in the development of the forecasting models. Two XGBoost models were developed: one excluding social media’s features and one including them. The model incorporating social media data outperformed the one without it. Furthermore, a feature selection process was applied, resulting in even more accurate forecasts. These findings suggest that integrating social media data can significantly enhance the accuracy of forecasting models for international student arrivals.