Bayesian modeling of student performance dynamics based on LMS interaction data

This study proposes a method for assessing college students' academic achievement and predicting their learning outcomes using Bayesian inference techniques. Furthermore, the Bayesian framework provides a new and more flexible approach to statistics. By fitting posterior probability distributions that characterize the uncertainty in information, our analysis can be more robust and more understandable. The experiments use a university's Learning Management System (LMS) data. We analyze student submission patterns and activity levels in detail. Using Markov Chain Monte Carlo (MCMC) simulations, a stochastic model was developed to view changes in learning behavior over time. Synthetic datasets were used to validate the model's predictions. The model also pinpoints critical parameters – such as submission intensity and switch points – influencing academic outcomes. The results demonstrate the effectiveness of Bayesian modeling for forecasting success at various learning stages. We also conclude with some practical recommendations about optimizing the curriculum and providing students with personalized support. This research incorporated the field of learning analytics by using real-world educational data to increase decision-making in higher education and then adding probabilistic methods.