Hybrid Time-series Representation for the Classification of Driving Behaviour

Abstract

The classification of driving behaviour is important for monitoring driving risk and fuel efficiency, as well as for providing a personalized view, or 'fingertip', of each driver, useful in driving assistance and car insurance industry. Intuitively, an aggressive driving style manifests itself in the long run, with distinct frequencies of occurrence for time-series patterns and critical events, such as accelerations, brakings and turnings. In this work, we consider a hybrid classification method, which employs both RNN-guided time-series encoding and rule-guided event detection. Histograms derived from the output of these two components are merged, normalized and used to train a standard perceptron to classify overall driving behavior as normal or aggressive. Experimental evaluation on a publicly available dataset of sensor measurements obtained for various drivers and routes, lead to the conclusion that both RNN-guided and rule-guided components contribute to the obtained classification accuracy. © 2020 IEEE.