Data Mining User Activity in Free and Open Source Software (FOSS)/ Open Learning Management Systems

Free and Open Source Software (FOSS)/Open Educational Systems development projects abound in higher education today. Many universities worldwide have adopted open source software like ATutor and Moodle as an alternative to commercial or homegrown systems. The move to open source learning management systems entails many special considerations, including usage analysis facilities. The tracking of users and their activities poses major technical and analytical challenges within web-based systems. This paper examines how user activity tracking challenges are met with data mining techniques, particularly web usage mining methods, in four different open learning management systems: ATutor, LON-CAPA, Moodle, and Sakai. As examples of data mining technologies adapted within widely used systems, they represent important first steps for moving educational data mining outside the research laboratory. Moreover, as examples of different open source development contexts, exemplify the potential for programmatic integration of data mining technology processes in the future. As open systems mature in the use of educational data mining, they move closer to the long-sought goal of achieving more interactive, personalized, adaptive learning environments online on a broad scale. DOI: 10.4018/jossp.2010010105 66 International Journal of Open Source Software and Processes, 2(1), 65-75, January-March 2010 Copyright © 2010, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. As projects and products, ATutor, LON-CAPA, Moodle, and Sakai differ in terms of technology, design, size, target audiences, organizational structures, funding models, software licensing choices, extendibility, and their pace of development. In no small way, analyzing why and how educational data mining techniques are being employed in these four open source contexts also involves understanding the underlying organizations and processes, not just the technologies and system designs. Data mining of web-based educational systems is also not particularly new. For more than a decade, exciting and visionary applications of educational data mining in web-based learning systems have been conceived, prototyped, and tested (Johnson, 1998). As Romero and Ventura show, much of the work to date has been carried out mainly within the lifecycle and setting of individual research and development projects with small populations of research subjects (Romero & Ventura, 2007). In contrast, the projects described in this paper involve data mining technologies and techniques being adapted to widely available open source learning management systems. For many reasons, traditional techniques for studying online users have proven inadequate when applied in systems of the scale and complexity seen in these widely adopted, open technology and open learning management systems (McGrath, 2008b). As an alternative for studying usage, educational data mining techniques are being embraced because the techniques excel at uncovering usage structure and access patterns within the enormous volume and variety of loosely structured user data generated in large systems. And while exploratory in nature, the four projects described here are not only for research and development. They lean more towards developing practical approaches to building and managing distributed capture and analysis systems for large web-based production deployments in production at many universities and colleges. The four projects involve fairly simple applications of data mining, from external usage analytics to association-rule based feedback, that only begin to tap the fuller potential educational data mining has shown in pure research prototype systems. Two are grant funded research experiments, but in all four cases practical needs have motivated the turn towards applying the algorithms of data mining and machine learning as a way of understanding users and usage.