Visualisation of learning management system usage for detecting student behaviour patterns

Abstract

Identifying \at-risk" students - those that are in danger of failing or not completing a course - is a crucial element in enabling students to achieve their full potential. However, with large class sizes and growing academic workloads, it is becoming increasingly dicult to identify students who require urgent and timely assistance. Ecient and easy to use tools are needed to assist academics in locating these students at early stages within their courses. A signicant body of work exists in the use of student activity data, e.g. attendance, performance, participation in faceto- face and online sessions, to predict overall student performance and at-risk status. This is often built upon the considerable amount of student data within learning management systems. Manual data collection, including surveys and observation, which introduces additional workload is often required to extract relevant data meaning that it in large classes it is prohibitively dicult to apply such techniques. In this paper, we introduce a framework for atrisk identication combining simple metrics, gathered from social network and statistical analysis domains, that have been shown to correlate with student performance and require slow amounts of manual data collection or additional expert analysis. We describe each of the metrics within our framework and demonstrate their usage. We use visualisation to enable easy interpretation of results. The application of our framework is demonstrated within the context of an advanced undergraduate computer science course.