AI-driven student profiling: a cross-industry review and future research directions on machine learning for admissions and retention

Higher education institutions are increasingly facing challenges in student recruitment and retention, necessitating the development of data-driven segmentation and profiling models.Traditional segmentation approaches rely heavily on historical data and often fail to capture shifts in applicant behaviour.In contrast, AI-based techniques enable real-time analysis of interactions, supporting more adaptive and personalised recruitment strategies.Across fields like marketing, finance, and e-commerce, researchers have shown that clustering, supervised learning, and hybrid ML models can meaningfully improve segmentation tasks.This article explores the application of such methodologies in the context of higher education.A systematic literature review (SLR) and a Weighted Sum Model (WSM) were applied to evaluate and rank the relevance and adaptability of segmentation techniques used in the analysed studies.Results indicate that K-Means and Random Forest are widely applied in student segmentation and performance prediction.At the same time, hybrid models, such as K-Means combined with decision trees, offer improved accuracy and interpretability.However, real-time behavioural analytics (e.g., Google Analytics) and natural language processing (NLP) methods remain underexplored.Overall, the findings suggest that AI-driven segmentation models can enhance student engagement, academic success prediction, and strategic enrolment management.Future work will focus on integrating real-time behavioural data, refining segmentation models, and improving predictive analytics for university admissions and retention strategies.