The development of in-vehicle telecommunication devices (telematics)-technology, wireless connectivity, machine-tomachine communication, and mobile applications powered the development of usage-based insurance tracking vehicle distance driven and driving behaviour. This paper investigates the added value of combining traditional rating factors with driving behaviour data obtained using telematics to improve automobile insurance risk management. Two classification techniques are used for investigating the claim frequency: (i) a classical Generalized Linear Model (GLM) with Poisson distribution for the expected number of claims, and (ii) a Bagging (Bootstrap Aggregation) GLM machine-learning technique. The empirical results suggest that the vehicle distance driven influences the probability of having a road accident and, thus, the cost of auto insurance coverage. This means that the use of telemetric data has the potential to improve risk management in insurance, facilitate price discrimination and reduce unintended cross-subsidies between policyholders.

--

Risk management,telematics,Usage-based insurance,Bagging-GLM,Price discrimination