A Comparative Analysis of Different Machine Learning Models for Classifying Student Achievement

Abstract

In the context of teaching and learning, evaluating and classifying student achievement is critical for determining the effectiveness of instructional methods. Categorizing students’ academic performance into groups such as “passed,” “failed,” “successful,” and “unsuccessful” provides valuable insights for tracking academic progress and improving instructional strategies. The use of Machine Learning (ML) models in such classifications enables more accurate and objective evaluations, particularly when dealing with large datasets. Therefore, this study aims to examine the accuracy of various ML models in classifying student performance. ML offers enhanced precision and objectivity by analyzing large and complex educational datasets. In this study, the classification accuracies of three machine learning algorithms—Naive Bayes (NB), Support Vector Machines (SVM), and Random Forest (RF)—were evaluated. The research compares the performance metrics of these models in predicting students' academic success and examines the results in detail. As such, the study adopts a descriptive survey design and has an applied nature. A dataset comprising 1,000 samples and variables such as ethnicity, parental education level, and mathematics achievement was used. The analyses were conducted using SPSS and R software. The findings reveal that the Random Forest model achieved the highest classification accuracy. The integration of ML models in education can contribute to improving educational quality by predicting student success, identifying risk of failure, and evaluating the effectiveness of instructional methods and materials.