Pedestrian Spatial Violation Analyses for Urban Roadways

Abstract

This study presents an investigation regarding the critical contributing factors in pedestrian spatial violations based on field observation of 15,090 samples at 14 roadway segments in Shanghai, China. A violation prediction model was applied to predict the impacts of roadway geometry design and traffic on the number of violations, and a real-time pedestrian violation prediction model was used to predict whether a pedestrian would spatially violate. For the violation prediction model, a Bayesian Poisson-lognormal model was used, and for the real-time pedestrian violation prediction model, a Bayesian logistic regression model was adopted. Then, random forest was employed to rank the importance of factors that are significant in violation prediction. The results showed that the presence of median, land use type, and number of lanes are the most significant variables in spatial violation. The findings of this study can provide a basis for traffic practitioners, researchers, and authorities to analyze the reasons for pedestrians’ spatial violations and develop guidelines for crossings design.