Utilizing an Artificial Intelligence-Based Approach to Evaluate Student Academic Performance through the Application of an Ensemble Algorithm

Resumo

This study addresses the use of machine learning to identify a pattern for predicting student academic performance in the Philippines. The data used in this study comes from the UCI Student Performance dataset. The researcher used Multinomial logistic regression, Long Short-Term Memory, and eXtreme Gradient Boosting to create an ensemble model in which all three algorithms work together to solve the problem, in this case, finding a pattern for student performance, and uses 10-fold cross-validation for robust model training. The dataset is divided into four categories for Student Performance, Portuguese language dataset, Math dataset, joint dataset, and merge Dataset. The results demonstrate that the Ensemble model achieves high predictive performance for student-combined-join (accuracy of 86.5), student-merged (accuracy of 85.4), student-mat (accuracy of 86.5), and student-por (accuracy of 84.8). The results show that the created model could further enhance students' performance by using predictive models that help students, schools, and parents identify shortcomings in students and curriculum outcomes.  The model exhibits robust performance across all evaluated systems, suggesting that thoughtful model selection and combination yield accurate classifications across various metrics. This indicates not only the effectiveness of individual models but also the importance of their synergistic interaction in improving overall classification accuracy.