Finding Driver Styles on Driver Behavior Data with Unsupervised Learning
Finding Driver Styles on Driver Behavior Data with Unsupervised Learning, Proc RECPAD Portuguese Conf. on Pattern Recognition - RecPad RECPAD, Leiria, Portugal, Vol. , pp. - , October, 2022.
Digital Object Identifier:
In most countries, insurance against civil liability for vehicles is mandatory. Usually, insurance companies set vehicle insurance rates according to static variables, such as the age of the driver, the number of years one holds a driving license, and the driving history. These variables may not reflect the everyday behavior of the driver on the road, thus ending up by penalizing young good drivers. Moreover, an automatic driver style identification has many useful applications such as fleet management or
promoting good drivers. In this paper, we follow a pay-as-you-drive approach, to devise a driver style identification strategy, based on real-time driver behavior data. From data records with the trips from different drivers, we build a dataset. Then, we apply unsupervised machine learning techniques that are able to identify some distinct driver styles.