par Sente, Zoé 
Président du jury Deelstra, Griselda
Promoteur Trufin, Julien
Publication Non publié, 2021-01-19
Président du jury Deelstra, Griselda
Promoteur Trufin, Julien
Publication Non publié, 2021-01-19
Mémoire
| Résumé : | Which amount of money should an individual pay to cover its driving behavior against an adverse event? How could a coverage provider determine the inherent risk caused by its portfolio of car owners?For many years, the insurance industry has attempted to design innovative solutions to answer these questions. Indeed actuaries proposed models based on standard variables such as the driver's age or sex to predict the client's likelihood of entering an adverse event. However, those methods immediately presented various imperfections that can result in unfair situations for some customers. For example, a young driver can be an excellent driver, or a woman can show exemplary driving behavior. Nevertheless, both individuals end up paying an improper high premium.Later, thanks to improving technology, data science introduced innovative solutions to align the individual features and the premium paid. Indeed, the big data method's ability to operate with a high number of data created new opportunities for the industry. Today, those new models enable us to use various variables to capture the policyholder features and predict their likelihood to encounter an adverse event. For instance, we can use simultaneously the drivers' age, location, the car's weight, speed, and many other variables that can make a drive unique compared to the others. Nonetheless, as precise as those models' outputs can be, they do not give us any clue on the policyholder's driving behavior. Living in an urban area, being young, and driving a fast car does not necessarily make someone a terrible driver. Today, telematics appears as an innovation that could correct the flaws seen in previous models. The idea is to let the driver tell us how much he should pay to be covered. Thanks to a small device installed in the client's car, insurers can retrieve their clients' real driving data and compute the final premium on the drivers' behavior. In this scenario, individual personal features, such as age or location, are no longer relevant. The ability to follow the road code and demonstrate responsible driving skills will determine the premium.This technology presents essential questions and challenges. Consequently, This paper is meant to address the following issues: First, the nature of the data collected is quite unique and can not be included as such in existing models. Can those data be integrated in a pricing model and result in relevant outcomes?Second, a terrible driver can drive for many years without any claim. Can the data collected by the telematics predict the likelihood of a policy holder to enter in an adverse situation |
