Africa Education Review

Educational Data Mining Approaches for Predicting Student Performance and Retention

•        Educational Data Mining (EDM) finds wider use in the prediction of student performance and improving retention in learning institutions. EDM can be used to detect students at risk by analyzing academic records, learning behaviors and patterns of engagement and provide teachers with timely interventions. Key methods encompass classification models to predict discrete events such as pass or fail, regression-based models to predict grades, clustering models to cluster students with similar learning patterns and Deep Learning (DL) models that can model sophisticated relationships within large datasets. Furthermore, some of the main determinants of performance usually involve the previous academic background, involvement in coursework and regular use of learning platforms. Recent approaches, including Explainable Artificial Intelligence (XAI) and ensemble learning, improve accuracy in prediction and transparency to allow educators to know why a student may underperform. In addition, the insights contribute to individualized learning, specific interventions and proactive retention plans, making EDM an essential tool in modern education.  
•         Additionally, the introduction of EDM into institutional practice can change the manner in which educators facilitate student achievement. Targeting the right predictors and using sophisticated Machine Learning (ML) algorithms, institutions will be able to adopt early warning systems that will recognize the at-risk students before the performance deteriorates. Engagement measures, assignment completion trends and behavioral indicators can give crucial data to enhance predictive reliability, whereas explainable models can be used to give actionable information to instructors. EDM also guides curriculum design by identifying the gaps in knowledge and facilitating the provision of specific support to various learning needs. However, these methods will work only with quality data and prudent interpretation. Moreover, the application of EDM in predicting student performance and retention allows institutions to shift from reactive measures to data-informed approaches for supporting students.
•         In this research, we provide a forum for researchers and educators working on EDM to discuss approaches for predicting student performance and improving retention in learning institutions. Research work related to the application of classification, regression, clustering, DL, ensemble learning or XAI for identifying students at risk or improving the accuracy of predictions is especially encouraged. Research focusing on academic performance, engagement, behavioral and curriculum-related aspects or offering recommendations for making interventions and educational practices data-driven are invited to contribute to the development of EDM for supporting student success.