Enhancing collaborative filtering-based recommendation systems through Ensemble Learning in education

This study investigates the effectiveness of ensemble learning approaches in enhancing collaborative filtering-based recommendation systems in educational settings. Significant improvements in performance indicators were obtained after merging predictions from three different models - Singular Value Decomposition (SVD), Non-negative Matrix Factorization (NMF), and CoClustering - using ensemble learning approaches. Bagging, boosting, and stacking techniques led in significant reductions in Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) as compared to individual models. Specifically, the stacking strategy produced the most promising results, with an RMSE of 0.1572, an MAE of 0.0445, and an excellent R-square value of 0.9492. These findings highlight the efficacy of ensemble learning techniques in refining recommendation systems for educational applications, demonstrating its potential to improve tailored learning experiences.