Academic Performance Prediction Model Using Classification Algorithms: Exploring the Potential Factors

Abstract

The student's performance has become the focus in higher education institutions. The ability to predict students' performance is beneficial to improve their achievement and the learning process. However, producing a prediction model for academic performance becomes challenging when an educational dataset contains various data. Many researchers have widely explored this kind of research, but many features should be investigated to affect students' achievement. Finding the potential factors influencing students' performance helps enhance students' quality. These factors will assist an institution plan a strategy for improving students' performance. This research proposes a classifier model to predict students' academic performance and define the factors influencing the performance by considering 14 attributes from demographics, learning styles, and educational background. The model development employs seven machine learning algorithms, and the best model will be selected. The factors that influence academic performance will be revealed from that model. The dataset was collected by conducting a survey at UiTM Seremban involving 233 students from Science and Technology and Social Science Streams. The Random Forest Tree produced an accurate result with the simple rules to be interpreted. The model also showed four attributes: qualification before tertiary education, SPM result, Seniority and gender positively impacting academic performance. Some factors that did not influence their performance were their parents' academic background and hometown.