Risk assessment of driver emotion using a data-driven Bayesian Network: Insights from a naturalistic driving study

Abstract

Introduction: Driving in a state of high emotions, particularly anger, can impair driving abilities and increase the risk of crashes. Understanding the influencing factors on emotions and how emotions impact safety–critical events (SCEs) is crucial for enhancing traffic safety. This study aims to investigate the relationship between emotions and SCEs on the road considering various contributing factors such as driver demographics, traffic conditions, and road geography. Method: Through a data-driven Bayesian Network (BN), we validate the relationships among these factors and assess their joint impact on SCEs. By employing domain expert knowledge and a hybrid structure learning approach, we construct a BN structure that effectively captures the dependencies among different variables. Results: Clear dependency relationships are identified, with traffic flow directly influencing emotion, and factors like traffic density, secondary tasks, emotion, and driving errors significantly affecting SCEs. Through BN inference, we explore the marginal and joint effects of various factors on SCEs, highlighting the increased SCE risk associated with emotions, driving errors, high traffic density, and secondary tasks. Practical application: The identification of high-risk scenarios provides valuable insights for targeted intervention and mitigation efforts. This study contributes to a deeper understanding of the relationship between emotions and SCEs, offering methods for enhancing road safety measures and accident prevention strategies.