Investigation of Pedestrian Collision Severity Patterns at Highway-Railway Grade Crossings Collisions Using Integrated Machine Learning and Bayesian Hierarchical Models

This study aims at developing a solid understanding of the contributing factors to pedestrian fatal and injury collisions at highway-railway grade crossings (HRGC), along with the impact of different warning devices that are commonly used at HRGCs. The study utilized an integrated Machine Learning and Bayesian models to analyze the United States HRGC collision database between 2009 and 2018. First, the categorical principal component analysis (CATPCA) technique was utilized to reduce the complexity of the collision database. Second, a popular Machine Learning model (the Latent Class Analysis clustering method) was applied to cluster the collisions into a set of clusters, in which the different factors impact pedestrian collision severity in a similar manner. The goal is to overcome the inconsistency that is associated with the impact of some factors on collision severity and understand how different factors impact collision severity differently under different HRGC conditions. Finally, the Poisson-Lognormal model was developed to assess the impact of each explanatory variable on pedestrian fatalities and injuries in each cluster. The results demonstrate the association between different factors and the collision severity in each cluster and attempt to explain the inconsistency associated with the impact of some factors, such as weather conditions and pedestrian traits, on collision severity. The results also highlighted the conditions at which the different types of countermeasures and warning devices are most effective and the circumstances that limit their benefits. The results can aid engineers and planners to develop specific policies and designs to enhance pedestrian safety at HRGCs.

Language

  • English

Media Info

  • Media Type: Web
  • Features: Figures; References; Tables;
  • Pagination: 16p

Subject/Index Terms

Filing Info

  • Accession Number: 01763838
  • Record Type: Publication
  • Report/Paper Numbers: TRBAM-21-00826
  • Files: TRIS, TRB, ATRI
  • Created Date: Dec 23 2020 11:12AM