Understanding distracted driving patterns of ride-hailing drivers from multi-source data: Applying association rule mining

This study aims to explore the association between distracted driving (including cognitive, visual, operational, and auditory distractions) and multi-source features of ride-hailing drivers, and to discover frequent patterns of distracted driving. To achieve this, a natural driving experiment was conducted, and an association rule mining ('Apriori' algorithm) was used to uncover hidden rules between distracted driving and the multi-source features (including emotion, valence, arousal, driving tasks, cumulative driving hours, velocity, longitudinal acceleration, lateral acceleration, heading rate, presence of intersections, traffic status, and driving time of day). Results indicate that distracted driving is prevalent among ride-hailing drivers in specific scenarios, including trips without passengers, non-intersection sections, driving for over 4 h, and congested traffic conditions. The emotional state of drivers has also been found to have an interesting association with distracted driving. For instance, cognitive and operational distraction were highly associated with positive driving emotions, but auditory distraction caused by passenger interference was highly associated with negative driving emotions. Moreover, there are variations in distracted driving patterns across different categories. These findings can help ride-hailing platforms develop more scientific and effective distracted driving monitoring and prevention strategies.