King's College London

Public transport is perhaps the most significant component of the contemporary smart city currently being automated using sensor technologies that generate data about human behaviour. This is largely due to the fact that the travel associated with such transport is highly ordered. Travellers move collectively in closed vehicles between fixed stops and their entry into and from the system is unambiguous and easy to automate using smart cards. Flows can thus be easily calculated at specific station locations and bus stops and within fine temporal intervals. Here we outline work we have been doing using a remarkable big data set for public transport in Greater London generated from the Oyster Card, the smart card which has been in use for over 13 years. We explore the generic properties of the Tube and Overground rail system focusing first on the scale and distribution of the flow volumes at stations, then engaging in an analysis of temporal flows that can be decomposed into various patterns using principal components analysis (PCA) which smoothes out normal fluctuations and leaves a residual in which significant deviations can be tracked and explained. We then explore the heterogeneity in the data set with respect to how travel behaviour varies over different time intervals and suggest how we can use these ideas to detect and manage disruptions in the system.