Self-Regulation and Student Success : Building a Data-Driven Approach to Understand Student Learning Behaviours in Computing Education

The shifts toward student-centred teaching and the adoption of novel digital tools have expanded personalised learning opportunities, but they have also increased the responsibility for managing their own learning process. In response, education organisations have sought to better understand how students learn and where or when they struggle in order to provide timely support. Research on student learning behaviours and factors affecting academic performance exists, but some open areas remain, such as the use of behavioural pattern recognition as a tool for automatically mapping learning behaviours, the difficulty of explaining performance predictions generated by machine learning approaches, and debates about using self-reports versus analytics to measure behaviour. This dissertation presents a data-driven approach to understanding learning behaviour and linking it to academic success. The approach uses learning analytics to capture observable behaviours and interpret them in terms of dropout risk. The approach also combines trace log data with self-reports to enhance the observations with students’ personal experiences and aptitudes. Three studies were conducted to develop and review the approach within STEM and computing education. The first study introduced a tool using trace logs and self-reports in a mathematics classroom, identifying common behaviour types and associating them with self-reported self-regulation. The second study modelled student dropout in an introductory programming course, producing a dropout prediction model. The third study expanded the two studies by combining dropout prediction with process mining techniques and open-ended self-reports to identify distinct learning strategies associated with high and low dropout risk. Key findings include the validation of the approach and the identification of learning patterns and their connection to academic success in the studied contexts. The approach also demonstrates how combined use of learning analytics and self-reports can provide actionable insights into learning behaviours. The results can inform future interventions and tool development, and be used to educate teaching staff. Keywords: self-regulated learning, learning analytics, student dropout prediction