Student dropout continues to be a major concern in Higher Education, with far-reaching consequences for both individuals and Institutions. There is a growing interest in data-driven solutions within the academic community. This literature review explores 71 studies published between 2018 and 2025. This review intends to find the factors that contribute to student dropout and the algorithms used to predict student attrition. The analysis examines dataset characteristics, the modeling techniques applied, and the metrics chosen to evaluate performance. The review looks at the types of data used, the size and source of the datasets, the modeling techniques applied, and the metrics chosen to evaluate performance. Some traditional algorithms were commonly used in studies where transparency and ease of interpretation were the key. Some ensemble algorithms stood out for their strong performance in complex or imbalanced datasets achieving strong performance. Deep Learning approaches also showed promising results, having potential when applied to large behavioral datasets, although raising questions around explainability. This review aims to provide researchers and Institutions with practical insights into building more accurate, fair, and actionable dropout prediction systems.
Duro et al. (Thu,) studied this question.