Session events for a computer game

We study early player behaviour in a live computer game, focusing on the first hours and days of new-player activity, and examine how behavioural trajectories relate to long-term retention. Using detailed event logs, we apply sequential pattern mining methods such as PrefixSpan to extract recurring subsequences that capture onboarding pathways and potential friction loops. The project explores how to represent low-level events as higher-level behavioural primitives, how to incorporate session/time constraints, and how to evaluate pattern stability across player cohorts. Discovered patterns will be tested for predictive value in retention modelling and assessed for interpretability and actionability. The project offers hands-on experience with real-world sequential data, scalable analysis, and applied ML research suitable for an MSc thesis.