<![CDATA[Background: Bayesian Knowledge Tracing (BKT) is widely used in adaptive learning systems to model students’ knowledge mastery over time. However, conventional procedural implementations of BKT often suffer from limited modularity, maintainability, and integration capability, making them less suitable for modern adaptive learning environments. Objective: This study aims to design and evaluate an Object-Oriented Programming (OOP)-based BKT framework to improve software modularity and predictive performance in adaptive learning systems. Methods: The proposed framework applies OOP principles by decomposing the BKT model into modular classes, including StudentModel, Skill, and BKTUpdater. The framework was implemented in Python and evaluated using student interaction datasets through a 5-fold cross-validation approach. Performance was assessed using Accuracy, Area Under the Curve (AUC), and Root Mean Square Error (RMSE). Results: Experimental results demonstrate that the proposed OOP-based BKT framework outperforms the conventional implementation. The model achieved an improvement of approximately 1.53% in Accuracy and 1.06% in AUC, while reducing RMSE by 2.85%. Although OOP does not alter the mathematical formulation of BKT, it enables more structured parameter management, reduces implementation inconsistencies, and facilitates systematic experimentation, which indirectly contributes to enhanced predictive performance. Conclusion: The integration of OOP principles into BKT development improves both software architecture quality and model reliability in adaptive learning systems. The proposed framework provides a scalable, maintainable, and extensible foundation for future adaptive learning technologies while achieving better predictive performance than conventional BKT implementations.]]>
Copyrights © 2026
