Learning About Student Performance From Moodle Logs in a Higher Education Context With Gaussian Processes

Abstract

Learning Management Systems (LMS) serve as integral tools for executing and evaluating the educational journey. As students engage with the platform, LMS consistently collect valuable data on their learning progress. This study employs statistically-driven methodologies to gain insights into student performance, focusing exclusively on data derived from Moodle LMS, a widely adopted platform across educational institutions globally. In particular we take advantage of the Gaussian Process regression method in order to predict the marks of the students given their activity in Moodle, achieving up to 0.89 R. Besides the use of an advanced kernel, the Automatic Relevance Determination (ARD), allows us to analyse which variables are more relevant when predicting the continuous mark and which are relevant to predict the final mark. Analysing logged data spanning various subjects and degrees, our findings reveals the significance of the frequency of interactions with the LMS as a robust indicator of student performance. This observation suggests the potential utility of interaction metrics as effective measures for monitoring and assessing students’ ongoing learning trajectories. The implications of these results can extend to informing educational strategies and interventions to enhance student outcomes within the higher education field.