Mining Educational Data to Predict Student Performance Using Ensemble Learning

Abstract

The digital transformation of educational systems has generated vast quantities of data, providing an unprecedented opportunity to leverage machine learning for academic analytics. This paper explores the application of ensemble learning methods, specifically Random Forest, XGBoost, and a soft voting classifier, to predict student academic performance. The objective is to develop predictive models that can identify students at risk of poor academic outcomes early in the academic cycle. Such predictive systems can empower educators and administrators with insights to design timely intervention strategies. Using a dataset derived from student academic records, demographic profiles, and behavioral metrics, the study evaluates model performance based on accuracy, precision, recall, F1-score, and ROC-AUC metrics. The ensemble learning approach demonstrates superior performance over traditional individual classifiers. The paper discusses the implications of these findings for educational policy, data governance, and the ethical use of AI in educational environments.

Keywords: Educational data mining, student performance, ensemble models, predictive analytics