Support Vector Machine and Boosted Regression Tree as Effective Data Mining Tools for Predicting Students' Academic Performance

• Published in: IEEE International Conference on Emerging & Sustainable Technologies for Power & ICT in a Developing Society (Print) pp. 1 - 5
• Main Authors: Oparauwah, Nnaemeka M., Njoku, Donatus O., Onyemauche, Uchenna C., Okpalla, Chidimma L., Odii, Juliet N., Jibiri, Janefrances E., Okolie, Stanley A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26.11.2024
• Subjects: Academic performance; Accuracy; Boosted regression tree; Data analysis; Data mining; Data models; Humanities; Intelligent systems; Machine learning; Mathematical models; Prediction; Regression tree analysis; Support vector machine; Support vector machines
• ISSN: 2377-2697
• DOI: 10.1109/NIGERCON62786.2024.10927303

This paper reports the application of data mining techniques; Support Vector Regression (SVR) and Boosted Regression Tree (BRT) in prediction of students' academic performance in schools. The experimental approach in this study compared the performances of the two data mining techniques to ascertain the best performing model on an academic dataset. Publicly available dataset was used for this study and was retrieved from UCI Machine Learning Repository, Center for Machine Learning and Intelligent Systems. The data was collated from school reports and constructed questionnaires from two public secondary schools in Portugal regarding students' performance in two distinct subjects: Mathematics (mat) and Portuguese language (por). Data analysis was conducted in two schemes, first with the complete feature sets and then with the reduced feature set (containing 5 selected best performing features). Evaluation results showed that the BRT model outperformed the SVR model with 0.97% and 0.96% accuracy on the complete and reduced feature sets respectively.