Advancing Forecasting Methods for International Student Applications Using Machine Learning: A Comparative Analysis and Scenario Exploration

Ruth Nevile, University of Liverpool
Francisco Rowe, University of Liverpool
Emilio Zagheni, Max Planck Institute for Demographic Research (MPIDR)

The UK is the second largest destination for international students in the world. However, recent declines in applications from EU member states and slowing of growth in applications from China pose challenges in its standing in the international education market. Reliable forecasts of future international student applications are crucial for understanding the trajectory of the UK higher education sector. In this paper, we leverage a unique dataset using applications data from the UK Colleges and Admissions Service (UCAS) and data on the country-level of influences of international students. We aim to enhance current forecasting methods by comparing the effectiveness of ARIMA time-series models, negative binomial gravity models, and an XGBoost machine learning approach. We then compute several ‘what-if’ scenarios to understand how applications of students in the UK will be affected by demographic, economic, and social changes. Through this work, we seek not only to improve the accuracy of forecasting models but also gain insights into the dynamic nature of international student mobility. Our findings have implications for researchers, institutional stakeholders, and policymakers, providing valuable guidance for strategic planning and decision making in the higher education sector.

Keywords: Digital and computational demography, International Migration, Spatial Demography, Comparative methods

See extended abstract.