Effective mining of data from online submission systems offers the potential to improve educational outcomes by identifying student habits and behaviours and their relationship with levels of achievement. In particular, it may assist in identifying students at risk of performing poorly, allowing for early intervention. In this paper we investigate different methods of following the development of student behaviour throughout the semester using online submission system data, and different approaches to analysing this development. We demonstrate the application of these methods to data from a junior computer science course (N=494) and discuss their usefulness in understanding the common behavioural strategies of students in this course and how these develop over time. Finally, we draw links between behaviour in weekly coding tasks and student performance in the final exam and discuss whether these methods could be applicable midway through the semester. [For the full proceedings, see ED592609.]