Analysing student behaviour in online courses offers valuable insights to educators. In this paper we present methods for analysing student activity sequences to identify temporal behaviour patterns. We propose a new clustering-based approach, which integrates symbolic sequence representation, sequence similarity computation and change point detection, and also provides an interpretable visualisation of the student progress using Markov models. We conduct an evaluation in the context of a large-scale online programming course and show the effectiveness of our approach. The results enable educators to understand student behaviour patterns over time, and gain insights into the common behavioral patterns and critical points in time where student behavior changes. Our approach is applicable to other courses and provides the foundation for future research on student activity sequence analysis.