An analysis of diverse computational models for predicting student achievement on e-learning platforms using machine learning

Abstract

Coronavirus disease 2019 (COVID-19) has led many colleges and students to use online learning. In educational databases with so much data, evaluating student development is difficult. E-learning is essential for egalitarian education since it uses technology and contemporary learning techniques. This review research found three ways for predicting online course performance: i) To choose the best features to raise student performance; ii) The most effective algorithms for transforming unbalanced data into balanced data; and iii) The best machine learning algorithms to predict online course performance. This study also offered insights into using hybrid techniques and optimization algorithms to educational data sets to improve student performance prediction. The utilization of data from independent e-learning products to enhance education today requires data processing to ensure quality. In addition to these techniques, our abstract highlights the effectiveness of hybrid feature selection methods like L2 regularization (Ridge) and recursive feature elimination (RFE) and ensemble learning models like random forest, gradient boosting, and AdaBoost. These approaches considerably improve prediction accuracy and tackle huge and sophisticated educational dataset challenges. Our work uses advanced machine-learning approaches to optimize e-learning settings and boost academic achievements in the shifting online education landscape caused by the COVID-19 pandemic.