ABSTRACT
The advancement of predicting the performance of student is fast growing area of interest in educational domain, with significant implications for educational strategy and public policy. Machine learning (ML) has proven effective in investigating student data to recognize those at risk of poor outcomes, permitting for early support and intervention. Nevertheless, the complexity of these classifiers often poses interpretability issues, making it difficult for educators and policymakers to comprehend the underlying factors driving predictions. This paper proposes PolySHAP to enhance the transparency and interpretability of ML classifiers in educational settings. Numerous models are employed to identify and explain at-risk group of students based on certain features including academic, demographic, and behavioral. Proposing the PolyFeature technique is employed to improve model accuracy. The results demonstrate that the ensemble models perform better than single models and also the SHAP values effectively and efficiently decompose predictions into feature contributions, making the models’ decisions interpretable for the development of policy. By connecting model insights to actionable interventions, this study provides architecture for data-driven educational policies aimed at enhancing student performance, giving room to equitable resource distribution and decreasing attrition. The investigations emphasize the requirement for a balance between explainability and predictive performance in constructing policies that support success of the students and equal resource allocation in educational domain.
References
- [1] Ahmar, A. S., & Rahman, A. (2018). Development of prediction model of student academic performance. Journal of Physics: Conference Series_, 1028(1), 012162.
- [2] Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. In _Advances in neural information processing systems_ (pp. 4765-4774).
- [3] Fan, X., & Sun, W. (2017). Research on improved K-means clustering algorithm for college students. Journal of Physics: Conference Series_, 887(1), 012022.
- [4] Gray, K. E., & Perkins, M. C. (2019). Early engagement data for predicting student performance and risk of failure. Computers & Education_, 136, 60-68.
- [5] Francis, J., & Babu, M. S. (2019). A survey on predictive analytics in higher education to enhance student performance. _Journal of Physics: Conference Series_, 1228(1), 012011.
- [6] Mason, C. M., McFarland, J. W., & Shorter, S. H. (2018). Predicting student attrition using neural networks and course grades. _Journal of College Student Retention: Research, Theory & Practice_, 20(2), 172-186.
- [7] Fan, J., & Sun, M. (2021). Academic achievement prediction in higher education through interpretable modeling. PLOS ONE, 16(3), e0247590. https://doi.org/10.1371/journal.pone.0247590
- [8] Serrano, R., & Casillas, J. (2024). Predicting Academic Success of College Students Using Machine Learning Techniques. Data Journal, 9(4), 60. https://doi.org/10.3390/data9040060
- [9] Albreiki, I., Zaki, N., & Zubi, M. D. (2021). Predicting Student Performance Using Data Mining and Learning Analytics Techniques: A Systematic Literature Review. Applied Sciences, 11(19), 9023. https://doi.org/10.3390/app11199023
- Albreiki, I. M., Zaki, N., & Zubi, M. D. (2021). Educational Data Mining Techniques and Applications: A Review. Journal of Educational Technology Systems, 50(1), 41-73. https://doi.org/10.1177/00472395211014205
- Coussement, K., & De Caigny, A. (2020). Predicting Student Dropouts in Online Learning Using a Logit Leaf Model: Decision Support for Student Retention. Decision Support Systems, 138, 113382. https://doi.org/10.1016/j.dss.2020.113382
- Gray, J., & Perkins, D. (2019). Predicting Academic Challenges Using Engagement Data and Machine Learning Models. Journal of Educational Data Mining, 11(3), 1-19. https://jedm.educationaldatamining.org/index.php/JEDM/article/view/392
- Alelyani, S., & Zubi, M. (2021). Machine Learning Approaches for Predicting Academic Performance: A Review. International Journal of Advanced Computer Science and Applications, 12(1), 23-29. https://doi.org/10.14569/IJACSA.2021.0120146
Download all article in PDF
