Predictive Analytics for Learning Performance in First-Year University Programming Course

The increasing demand for programming skills has highlighted the need for effective teaching strategies to support student success in programming courses. Despite significant advancements in learning analytics, predictive models explicitly tailored to programming courses remain underexplored. This research aims to develop a machine learning model to predict student performance in programming courses offered within IT programs by analyzing gender, type of activity (readings, coding exercises, assignments), and frequency of access to different activities. Our study utilizes log data of the asynchronous learning activities in the learning management systems of the students enrolled in programming courses. We employ machine learning techniques, decision trees, gradient boosting machines (GBM), and logistic regression to build robust predictive models. In this study, the decision tree model outperformed logistic regression (77.77%) and gradient boosting machine (GBM) (86.57%) by achieving the highest accuracy of 89.09% and excelling in predicting 'Poor' student performance with a recall of 90.67%, establishing it as the most effective model for this predictive analysis. The findings from this research offer actionable insights for educators, enabling early intervention for at-risk students and developing tailored teaching strategies to enhance student performance through strategically provisioning the learning materials in programming courses. This study contributes to the growing knowledge of learning analytics and provides a foundation for future research in predictive modeling for diverse educational contexts.