A Data-Driven Approach for Customized Pay-As-You-Drive Insurance Premiums

Insurance companies have recently started to adopt usage-based insurance policies to provide insurance premiums according to the customer's driving behavior. Risk models are proposed based on the driver's behavior, driving history, and telematics data. These models do not consider exogenous factors, such as the geographical context, weather information, and driving events. This paper presents a data-driven approach for personalized premiums for car insurance by integrating endogenous and exogenous factors to calculate journey risk. The proposed algorithm uses accident, casualty, and vehicle data sets to calculate a baseline risk value for a given route. A final risk value is calculated by adding weather and journey-specific risk factors. The algorithm creates Voronoi regions to associate weather observations with geographical locations and analyses accident data to assign risk values to individual junctions and roads. The overall journey risk is calculated based on weather and accident risks, refined using metrics for the time of route or accident and journey purpose. The algorithm provides a simple and flexible way of calculating journey risk, considering multiple factors that can be used to provide truly personalized insurance premiums. The proposed approach has the potential to benefit insurance providers, regulators, and drivers by improving the accuracy of risk assessment and promoting safer journeys.