Predicting Freshman Recruitment Rates: A Comparative Analysis of Metropolitan and Non-Metropolitan Universities in South Korea

South Korea’s declining school-age population has intensified competition among universities, particularly in freshman recruitment, with non-metropolitan institutions facing disproportionate challenges. This study investigates regional disparities in recruitment rates by applying a range of statistical and deep learning models—including Generative Adversarial Networks (GAN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Autoencoders, and Transformer architectures—to predict freshman enrollment outcomes. These models integrate conventional institutional indicators such as scholarship amounts, graduate employment rates, and internationalization metrics, alongside newly emphasized variables like professors’ research performance. Findings demonstrate that including faculty research data significantly enhances model accuracy and predictive power, with Transformer-based models consistently outperforming others. The results underscore the methodological advantage of AI-driven modeling in educational analytics and offer strategic insights for policymakers: strengthening academic research capacity, especially in non-metropolitan universities, may serve as a vital policy lever for improving regional recruitment balance and sustaining higher education competitiveness.