Investigating risky driving events in urban corridors: A study with large-scale trajectory datasets

Ensuring corridor safety is essential for efficient transportation and minimizing economic losses. This study analyzes risky driving events—acceleration, deceleration, and overspeed—using over 20 million vehicle trajectories from Nanjing. The study investigates the impact of corridor geometry, traffic, and built environment on these events. A random parameters multivariate Poisson-lognormal (RP-MVPLN) model was developed to address unobserved heterogeneity and event correlation, with a traditional multivariate Poisson-lognormal (MVPLN) model used for comparison. Results show that the RP-MVPLN model outperforms the traditional model, highlighting the importance of accounting for unobserved heterogeneity. Findings indicate that different factors influence risky driving events in distinct ways. Geometric features like roadside parking and road length significantly impact all events. Higher traffic demand affects all behaviors, while speed limits…