The objective of this paper is to detect and analyze risky driving behaviour characteristics on the basis of smartphone data, with focus on key risk indicators, namely the number of harsh driving events and the use of mobile phone while driving. Driving behaviour analytics data from a naturalistic driving experiment are exploited in this research recorded by smartphone devices. The driving indicators that are collected include distance travelled, speed, accelerations, brakings, turnings, cornerings, and related ‘events’ in the form of harsh maneuvers (e.g. harsh acceleration, braking, etc.), as well as mobile phone use. One hundred drivers participated in the designed experiment during a 4-months timeframe and a large database of 18,850 trips was built. The results of this research reveal that distraction originating from smartphone usage has a serious impact on the number of harsh events that occur per kilometer and subsequently on the relative crash risk. Furthermore, mobile phone use while driving may be accurately “detected” by smartphone sensors data in more than 70% of cases.
|Tags||big data, driver behaviour, machine learning, naturalistic driving, statistical modelling, telematics|