On-Time Student Graduation Prediction Modeling: A Comparative Analysis of Naive Bayes Algorithm and Other Data Mining Classifications

Abstract

Predicting the on-time graduation of university students is a crucial task in higher education institutions, enabling proactive support and improving institutional effectiveness. This paper presents a comparative analysis of several machine learning algorithms for predicting on-time graduation, with a specific focus on challenging the performance of the Naive Bayes (NB) algorithm. Although often used as a baseline model, the effectiveness of NB in the complex domain of educational data is frequently debated. We compare NB with MultinomialNB and Decision Tree (DT), both widely favored in recent literature. Using a public dataset containing students' academic records, we follow the CRISP-DM methodology, incorporating feature selection and SMOTE to address class imbalance. The models are evaluated using accuracy, precision, recall, and F1-score metrics. Our results show that while Decision Tree achieves the highest accuracy, Naive Bayes offers an appealing balance of performance, computational efficiency, and interpretability, making it a strong candidate for implementation in early warning systems at universities. This study provides empirical evidence on the role of Naive Bayes in the current landscape of educational data mining. The classification results show an accuracy of 0.82 for Naive Bayes, 0.81 for MultinomialNB, and 0.85 for Decision Tree.