Analysis of big data set of urban traffic data

Modern vehicles are increasingly capable of reporting location and status information in real time using GPSenabled on-board telemetry boxes which connect directly into a vehicle’s control and diagnostic systems. We perform an exploratory analysis of such data obtained from a vehicle operating in a large UK urban area. The primary objective is to devise informative summary statistics that allow different ”types” of vehicle activity to be identified and abnormal behaviour to be quantified. We use position, speed, time and engine status (ignition on/off) to organise the data into working units of increasing temporal duration. We apply hierarchical clustering methods to some functions of these working units to identify different types of vehicle paths in addition to quantifying how the periodic daily variation in traffic conditions in a modern city affects fleet movements and behaviour.