Predicting Students’ Academic Performance Using a Hybrid of Machine Learning Algorithms

Abstract

The research highlights the significance of student academic performance as a crucial factor in determining the success of educational institutions. Educational data mining (EDM) is employed to analyze educational data, aiming to explore student academic performance. The study proposes a novel, feature-rich model for predicting student performance, integrating backlog information and student grades identified as key aspects through data mining techniques. It demonstrates the feasibility of developing a predictive model with satisfactory accuracy rates, even when trained on a limited dataset. Additionally, the research identifies critical parameters such as student behavior, family education, and subject grade averages essential for constructing the model and presenting data. To determine the optimal model, the study evaluates various algorithms using important attributes. A diverse set of classifiers, including decision trees, support vector machines (SVM), and k-nearest neighbor (KNN), is employed to assess the model's efficiency. Moreover, ensemble techniques such as bagging, boosting, stacking, and random forest are utilized to enhance classifier performance. The research concludes by establishing a clear correlation between students' attributes (such as social interactions and absenteeism), past exam performance (G2), family education (mothers' education), and their final grades (G3). With an accuracy rate of 91.5%, the findings validate the effectiveness of ensemble approaches in improving prediction models for student academic performance