Modeling the Frequency of Pedestrian and Bicyclist Crashes at Intersections: Big Data-driven Evidence From Maryland

Abstract

This study leverages big location-based service data collected from mobile devices in 2019 to conduct a pedestrian and bicyclist safety analysis. Statistical models are estimated for pedestrian and bicyclist crash frequency at Maryland intersections using location-based service data as risk exposure data. The analysis is performed by employing prominent frequency modeling methodologies, including Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial regression techniques. The findings indicate that inclusion of big location-based service exposure data in the analysis improves the performance of the models. Further, the results suggest that key contributing factors to pedestrian and bicyclist crashes at Maryland intersections include the following: (i) intersection design- and traffic-related attributes, such as the number of intersection legs, presence of a traffic signal, and average level of traffic stress rating, as well as safety risk exposure measures, such as the average daily pedestrian, bicyclist, and vehicle volumes at the intersection; (ii) travel-related attributes, including public transportation and nonmotorized mode shares within the intersection’s census block group; (iii) land use and built environment attributes, such as road network density, activity density, and extent of walkability within the census block group; (iv) socioeconomic and sociodemographic attributes, including the percentage of low-income workers, households with no vehicles, African American population, and senior population within the census block group. The findings of the study show how big location-based service exposure data can be utilized to identify pedestrian and bicyclist safety risks and guide data-driven, evidence-based policy decision-making to improve the safety of vulnerable road users.