Abstract

Imbalanced student performance data in educational institutions is crucial for any machine learning prediction model. It affects the efficiency of classifiers and challenges the sampling methods for having a more significant number of features. The proposed model was designed to predict the student's performance by comparing the results with popular similar models. The Kaggle dataset having 386 rows and 33 columns of student data, was used in this study. In addition, the Portugal database was considered for experimental analysis, containing 394 students and 19 attributes. It proves that the Random Forest classifier yields the highest accuracy at 84.15% in handling imbalanced datasets. The results show that the SVM-SMOTE is higher accuracy as, 94.76%, than the other sampling methods in predicting the student's performance with various features.