Educational Data Mining for Predicting Academic Outcomes Using Ensemble Techniques

Educational data mining (EDM) has gained significant attention in recent years for its potential to improve educational outcomes by analysing student data to uncover underlying patterns. This field is essential as it offers institutions actionable insights to enhance teaching methods, student engagement, and contribute to their overall academic success. This research specifically focuses on developing a model that predicts student performance using behavioural data derived from an e-learning management system. Previous research in EDM has identified gaps in understanding the full impact of behavioural features on academic performance, and existing prediction models often lack accuracy when applied to novel student data. Furthermore, previous studies have indicated the need to integrate multiple behavioral metrics in a unified predictive framework to increase reliability and relevance between diverse student populations. To address these current challenges, this research proposes a novel student performance prediction model that leverages advanced behavioral features, such as engagement metrics and resource utilization, from Learning Management Systems, making it uniquely suited for current online learning environments. The proposed model incorporates decision trees, artificial neural networks, and closest neighbor k-nearest as base classifiers. Furthermore, ensemble methods including Bagging, Boosting, and Random Forest are utilized to refine predictive ability which are then stacked together to build the final model. The results have highlighted that the ensemble learning methods have outperformed traditional base classifiers with respect to the evaluation metrics including R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, the mean absolute error, the mean squared error and the root mean squared error. These findings resemble the importance of behavioral data in EDM and highlight the potential for refined ensemble approaches to make meaningful contributions to educational analytics and student success.