A Machine Learning Approach for Predicting Student Engagement with Online Learning Platform

Online learning platforms such as Coursera, Udemy, and edX etc. become pivotal in modern education by providing students with flexible and accessible learning opportunities. Despite this, predicting student engagement across these platforms remains a significant challenge, especially in balancing online learning with traditional academic responsibilities. This study examines the engagement patterns of 422 university and college students through machine learning techniques. The dataset was preprocessed to address imbalances using SMOTE and underwent feature selection via chi-squared tests. Various machine learning models were evaluated, with CatBoost achieving the highest accuracy of 96.45%. Key engagement predictors, such as quiz completions, assignment submissions, and participation in discussions, were identified. These findings will provide online learning platforms with a clearer understanding of student engagement, helping improve platform design and enable instructors to create more engaging content.