(The) study of large-scale digital tachograph analysis for machine learning-based aggressive driving behavior classification

Abstract

Traffic accidents are one of the main causes of fatalities in Korea. The fatality rate was 2nd highest rate with 11.3 deaths per 100,000 people among the OECD countries. To reduce accident, it is important to analysis driver’s aggressive driving behavior and support them with proper feedback. Recent advances in sensor technology facilitates close monitoring of driving conditions by using digital tachograph (DTG). However, the effectiveness was decreases because it is formulated based on identical standards. In this thesis, we propose the quantitative evaluation criteria for aggressive driving behavior from large scale DTG data based on machine learning. The feature of DTG is extracted using abrupt change point detection method and artificial neural network. Driving pattern evaluation criteria is developed by two-level clustering and using the developed criteria, showing the possibility of diagnosis system for single driver’s behaviors. This thesis proposed the process for analyzing large-scale driving records based on machine learning algorithms and provide improving driving safety management system by establishing concrete and quantitative criteria of aggressive driving behaviors.