Learning Analytics and Student Retention: A Longitudinal Study in Virtual Higher Education

Abstract

Learning analytics (LA) has emerged as an effective tool to address challenges such as student retention and academic performance in virtual higher education programs. This study aims to analyze the relationship between participation patterns in digital academic activities and the academic performance of online postgraduate students, as well as their connection to student persistence. Three key dimensions are examined: participation trends and clusters over time; the factors that predict academic performance and retention; and the potential of predictive models to personalize pedagogical strategies. A quantitative, longitudinal, and exploratory design was employed, using a sample of 393 students enrolled at a Colombian university in 2024. Data was collected from institutional systems and analyzed through mixed linear models, machine learning algorithms, and time series analysis. The academic activities evaluated included application projects, practical cases, exams, discussion forums, and overall participation. Results indicate that practical activities, such as application projects and case studies, yield the highest levels of engagement and have a significant impact on academic performance. In contrast, forums and exams show lower levels of interaction. The clustering analysis identified three distinct student groups based on participation patterns, enabling more personalized pedagogical strategies. The findings suggest that LA provides a solid foundation for designing targeted interventions aimed at improving student retention and optimizing academic performance in virtual higher education. This approach reinforces the value of data-driven educational personalization.