Profiling drivers' risky behaviour towards all road users

Demographics, crash records and self-reported driving behaviour have typically been used as the basis for building driver profiles of crash risk. These capture the most serious of crashes but underreport other events such as less severe crashes and near-crashes. Improved technology has allowed for the collection of more disaggregate data on day-to-day driving. In turn, this has the potential for use in more comprehensive risk assessments. However, isolating the influence of the driver on behaviour from behaviour influenced by external factors including the road environment can pose a challenge. This paper presents a framework and methodology for profiling drivers along multiple dimensions of behaviour and risk to the driver and other road users using empirical data. Using 8 million second-by-second GPS data observations collected from 106 drivers in Sydney over several weeks, this paper examines the effectiveness of this approach. The results indicate that over 90 per cent of drivers exhibit more variability in speeding, acceleration and braking behaviour between different road environments than within the same road environment. This analysis points to the potential for using more disaggregate data but also the necessity to control for temporal and spatial factors when studying driver behaviour. Building comprehensive driver profiles using the proposed framework has the potential to provide a different way of classifying drivers other than demographics or (rare) crashes.