Evaluation of the Recommender and Predictive Algorithm in Learning Management System

Abstract

This study evaluates the accuracy of a predictive algorithm and the effectiveness of a recommender system integrated into a Learning Management System (LMS). The LMS was used by 387 second-year Information Technology students taking Object-Oriented Programming. The model utilized demographic and academic data—such as age, gender, class schedule, and previous programming grades—as predictors to forecast student performance. Students identified as at-risk of failing were divided into two groups: one receiving targeted content recommendations from the LMS, while the other received no intervention. The study measured the model’s performance through a confusion matrix, showing an accuracy of 86% and a kappa value of 0.61, reflecting substantial agreement between predicted and actual outcomes. Furthermore, an independent t-test revealed a statistically significant improvement (p = 0.002) in the final grades of students who received recommendations, with an average increase of 2.01 points compared to those who did not. The findings highlight the practical benefits of integrating predictive analytics and personalized learning interventions into LMS platforms. The study underscores the system’s ability to provide meaningful support to at-risk students, improve performance, and reduce variability in academic outcomes. Future work will focus on refining the algorithm to better identify failure risks and optimize the distribution of educational resources.