Usage-based insurance, also known as pay as you drive, is a type of automobile insurance whereby the costs of motor insurance are dependent upon type of vehicle used, measured against time, distance, behavior and place.
One important requirement of UBI is the ability to quickly quantify or classify the behavior of drivers based on the telemetry data gathered during their recent drives. Insurance companies need this kind of classification to quickly feed the driving pattern (safe or reckless) through into user’s premiums.
Getting good labeled data is always a challenge, when building a machine learning model for illustrative purposes we have reduced the problem to categorizing drives into one of four risk categories (0-3), representing excellent, good, fair, and high risk drivers. This allows us to use a supervised learning model which is a good fit for this problem.
For an insurance company, this example can be expanded with data they already have to provide a more complex set of risk classes. Insurance companies will keep the history of these drivers, so they could easily define “excellent” or “risky” according to the accident and insurance rates on different drivers.
Want to know more? Check out the work that Ivan Judson has done on GitHub.