top of page
27.jpg

Driver Ranking system

Insurance company underwrite of policies

Insurance companies traditionally compute a driver's risk by the number of accidents he had in the last 3-5 years. 

This a naive approach that does not consider how fault he was it, how much time the driver is using the road, and how good a driver is.


In this project, we used IoT sensors from the car such as speed, accelerometers, GPS, and more. We were able to rank drivers from best to worst by their likelihood of being involved in an accident.

Overcoming noise in the data was the greatest challenge, we did it using unsupervised learning. 


Our system had a high correlation between its score and the number of accidents the driver took part in.

Increase the accuracy of ranking a driver

Improve service and reward good drivers

Risk mitigation with bad or new drivers

Man on the Wheel

Business Value for the Client

  • Basis for a new service to sell to insurance companies.

  • Completely new revenue stream from their data

bottom of page