Student Performance Prediction Using Machine Learning Algorithms

Education plays a vital role in building a productive life by fostering self-confidence and providing essential knowledge and resources. With the rapid advancement of technologies such as artificial intelligence, higher education institutions are increasingly integrating technological tools into traditional teaching and learning methods. Predicting students’ academic success has become an important research focus, as strong academic performance not only enhances a university’s reputation and ranking but also improves graduates’ employment prospects. However, modern educational institutions face several challenges, including analyzing student performance, maintaining high-quality education, developing effective evaluation strategies, and anticipating future educational needs. E-learning has emerged as a rapidly expanding and advanced mode of education, allowing students to participate in online courses and learning platforms. Technologies such as Intelligent Tutoring Systems (ITS), Learning Management Systems (LMS), and Massive Open Online Courses (MOOCs) utilize Educational Data Mining (EDM) to support automated grading systems, recommendation systems, and adaptive learning environments. Despite these advantages, e-learning environments remain challenging due to the limited direct interaction between students and instructors. Machine learning (ML) plays an important role in the development of intelligent and adaptive systems capable of performing complex tasks that often exceed human capabilities. ML algorithms are widely applied in fields such as cluster analysis, pattern recognition, image processing, natural language processing, and medical diagnostics. In this study, the K-means clustering technique, combined with the Davies–Bouldin index, was used to identify clusters and determine significant features affecting student performance. The research evaluated several classification algorithms, including Support Vector Machine (SVM), Decision Tree, Naïve Bayes, and K-Nearest Neighbors (KNN). The results indicate that the SVM algorithm achieved the highest prediction performance after hyperparameter tuning, reaching an accuracy of 96%. Parameter optimization significantly improved the accuracy of all four prediction models. Among them, the Naïve Bayes classifier demonstrated the lowest predictive performance, primarily due to its assumption of strong independence among features.