Comparison of Machine Learning Methods for Student Graduation Prediction

Abstract

Student graduation prediction has become an important topic in higher education because it supports academic quality improvement, institutional performance evaluation, and early intervention for students at risk of delayed graduation. This study aims to compare the performance of machine learning methods used for student graduation prediction through a literature review approach. The literature was collected from Google Scholar, focusing on studies published between 2018 and 2025. Fourteen relevant articles were reviewed, consisting of twelve empirical studies included in the comparative analysis and two review articles used as supporting references. The main algorithms discussed are Random Forest, Decision Tree (C4.5), and Naïve Bayes, with additional comparisons involving K-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Logistic Regression, and Neural Network. The findings indicate that Random Forest generally demonstrates stable predictive performance across many studies, while Decision Tree provides better model interpretability and Naïve Bayes offers computational efficiency. However, differences in dataset characteristics, feature selection, preprocessing techniques, and evaluation strategies significantly influence the reported performance of each algorithm. Therefore, no single machine learning method can be considered universally superior for student graduation prediction. The review also identifies GPA, completed credits, attendance, and active semesters as the most influential factors in predicting student graduation