Machine Learning in Educational Data Mining: Current Trends and Emerging Gaps in Predicting Student Performance

Abstract

The growing availability of educational data and advancements in machine learning (ML) have led to its widespread application in predicting student performance. However, current research remains fragmented, lacking an integrated understanding of knowledge structures, collaboration networks, and thematic directions. This study addresses this gap through a bibliometric analysis of literature focused on ML-based student performance prediction. The study contributes a comprehensive knowledge map that identifies major research themes, key contributors, and underexplored areas within educational data mining. Using 465 Scopus-indexed articles from 2005–2025, four bibliometric techniques were applied: descriptive analysis, collaboration mapping, keyword co-occurrence, and thematic mapping. Data were processed using Bibliometrix in R and visualized via Biblioshiny. Findings reveal a sharp increase in publication trends after 2018, with China, the U.S., and India as top contributors. Despite high output, international collaboration remains limited to certain clusters, while countries like Pakistan and Indonesia show high collaborative intensity. Keyword analysis highlights “student performance” and “machine learning” as core themes, while federated learning, contrastive learning, and algorithmic fairness are emerging gaps. Non-cognitive factors such as motivation and emotional engagement are also underrepresented in predictive models. In conclusion, this study offers a systematic overview of the field, outlining its evolution, key players, and future directions. The results provide valuable insights for designing predictive models that are accurate, ethical, and contextually appropriate for higher education.