DATA MINING–DRIVEN STUDENT PERFORMANCE PREDICTION: A SYMMETRIC UNCERTAINTY-BASED COMPARATIVE ANALYSIS OF DECISION TREE AND SUPPORT VECTOR MACHINE MODELS

Abstract

Student academic performance refers to the measurable educational outcomes achieved by students through their learning process, while education represents the systematic framework designed to facilitate knowledge transfer and skill development. The use of machine learning and data mining has brought a noticeable shift in how educational data is studied. These methods help researchers and institutions make sense of complex student information by turning it into insights that can actually be used. With these tools, it becomes easier to predict how students might perform, spot those who may need support at an early stage, and take steps that can genuinely improve learning outcomes as well as the overall performance of an institution. In this study, the results clearly show that machine learning can be very effective for educational prediction. Among the models tested, the Decision Tree performed the best, reaching an accuracy of 86.5%, which is considerably higher than the Support Vector Machine’s 76.4% accuracy. The Decision Tree not only achieved perfect precision (1.00) for high-achieving students but also delivered stable results across all categories. Overall, these findings highlight how powerful data-driven approaches can be in helping educational systems respond to individual student needs and support learning in a more informed and timely way.