Using geolocation data, POIs and drivers’ behavioral information to infer risk profiles in future mobility scenarios

Abstract

The development of new technology-based products has enabled the insurance world to offer new tools that grant a certain level of control over the driver and the services associated with mobility. Analyzing driver behavior through objective data in the connected car offers a plethora of opportunities to model the way someone drives and the services that can be offered accordingly. In this sense, this paper analyses the development of a risk evaluation and management system empowered with drivers’ geolocation data and their Points of Interests (POIs) visited, compared with prior drivers’ behavior profiling schemes based on historical claims. For that, this paper compares several supervised classifiers’ approaches trained with 40,500 instances of drivers’ data obtained from the Spanish Historical File of Automobile accident rate database (SINCO). Our results show that an SVM-based classifier outperforms other alternative models in terms of accuracy. Thus, a novel customer risk analysis model based on their geo-localization and visited locations (POIs) is presented. The results and implications of the new model based on drivers’ lifestyle rather than driving-style for driver’s health and environmental impact are assessed in the paper.