Student Behaviour Models for a University LMS

Learning Management Systems (LMSs) enable teachers and educational institutions to manage the organization of the courses offered and deliver courses in blended form, with LMSs offering support to in-person teaching, or fully online. LMSs, despite having been used for a long time, saw a dramatic increase in usage due to the Covid-19 pandemic; the purpose of this study is the analysis of student behaviour within the Moodle platform, by exploiting the user interaction logs as recorded by the platform itself. Two models are proposed to predict the final outcome of students’ exams based on their behaviour within the platform. The first model consists of a support vector machine, while the second model consists of an artificial neural network; both models were tested on two real-world data sets, delivering outstanding results in terms of accuracy, above 90% for some of the tested configurations.