Large Scale Driver Identification Using Driving Data

Abstract

We address a large-scale driver identification problem whose goal is to predict the driver who drives a car from driving data collected using GPS devices. In contrast with the prior researches considering at most a few dozen of drivers, we try to identify a huge number of drivers up to 10,000 drivers. The experiment shows our method can identify drivers more precisely than baseline method. We also show that temporal features are quite effective in the large scale driver identification and that speed and acceleration features also contribute to driver identification.