BEYOND ACADEMIC METRICS: PREDICTING STUDENT PERFORMANCE USING ENGAGEMENT AND CONTEXTUAL INDICATORS

Abstract

This study investigates the role of engagement and contextual indicators in predicting student academic performance using machine learning techniques. Traditional academic metrics such as attendance and grades often fail to capture behavioral and psychological aspects of learning. Using a dataset of 177 student records, this study evaluates Logistic Regression, Decision Tree, and Random Forest models to analyze the influence of engagement-related factors on academic outcomes. Experimental results show that psychological engagement is the most influential predictor of performance. Logistic Regression and Random Forest achieved the highest prediction accuracy of 83%. The findings highlight the importance of incorporating engagement indicators into modern educational analytics systems.