Enhancing Prediction of Student Performance Based on Semi-Supervised Data Augmentation Approach

Abstract

– The problem of students dropping out of school is complicated, especially for information technology majors. One of the critical goals of contemporary institutions is to deliver quality education while simultaneously lowering academic failure rates. A university’s early identification of low-performing students may improve their academic performance, eliminate academic delays, and lessen the likelihood of academic failure. This study proposed a semi-supervised learning strategy to enhance the dataset for maximum accuracy in forecasting student dropouts. The proposed approach was evaluated and compared to various well-established methods, such as support vector machine (SVM), K-nearest neighbors (KNN), ensemble, and bilayer neural network (BNN). During a single academic year, the dataset was collected from the student management system at a private university. The dataset included grade point averages of 13 courses, the job status of students’ parents, and the geographical locations of the students enrolled in the computer technology program. The study’s findings suggest that when there is a shortage of appropriate training data, dataset improvement may enhance the classifier accuracy. According to the results of the classification technique, the proposed methodology attained a maximum degree of accuracy of approximately 98%.