Bayesian hierarchical spatial count modeling of taxi speeding events based on GPS trajectory data

Area of Science:

• Traffic Safety Research
• Transportation Engineering
• Spatial Analysis

Background:

• Speeding behavior, particularly serious speeding, is prevalent among taxi drivers due to extensive road exposure.
• Existing research on taxi speeding primarily focuses on driver demographics and operational factors, with limited exploration of road characteristics and spatial influences.

Purpose of the Study:

• To investigate the impact of 10 road characteristics and two spatial effects (spatial correlation and heterogeneity) on taxi speeding frequency.
• To determine the optimal statistical model for analyzing taxi speeding behavior.

Main Methods:

• Utilized taxi GPS trajectory data from a Chinese metropolis to identify speeding events.
• Developed and compared four Bayesian hierarchical count models (Poisson and negative binomial distributions) to assess contributing factors.
• Identified the Bayesian hierarchical spatial Poisson log-linear model as the most suitable for analyzing both total and serious speeding frequency.

Main Results:

• Drivers tend to speed more on long, multilane roads with median strips, non-motorized vehicle lanes, bus-only lanes, viaducts, or road tunnels.
• Low speed limits and work zones also correlate with increased speeding. Serious speeding is further influenced by speed cameras and one-way road organization.
• Both spatial correlation and heterogeneity significantly increase speeding events, with spatial heterogeneity having a more critical impact.

Conclusions:

• Road design and traffic management features significantly influence taxi speeding behavior.
• Spatial effects play a critical role in understanding and mitigating speeding incidents among taxi drivers.
• Findings provide valuable insights for developing targeted interventions to reduce speed-related crashes in urban taxi operations.