Applying model-based recursive partitioning to improve pedestrian exposure models to support transportation safety analyses

Abstract

Pedestrians are among the most vulnerable road users in urban areas, and their safety is a growing concern for transportation planners and engineers. Pedestrians are at disproportionately high risk for injuries or fatalities in crashes with motor vehicles, highlighting the critical need to address their safety. To address the dangers urban pedestrians face, the relationship between pedestrian crashes and their contributing factors must first be understood. One way to do this is to use statistical models relating pedestrian crash frequency with quantifiable contributing factors, such as land use, demographics, and roadway characteristics. Perhaps the most important of these factors is pedestrian exposure, which is often difficult to obtain because pedestrian volumes are not as widely available as vehicle volumes. Since pedestrian volumes are not available across an entire network, they are often estimated using statistical models—for example, negative binomial (NB) regression—rather than being directly observed. These models are typically a ‘one-size-fits-all’ approach, applying the same model to estimate pedestrian exposure across the entire network. However, relationships between pedestrian exposure and explanatory features—such as population, infrastructure design, and land use context—might differ significantly with respect to the context of an individual location. To address this issue, this paper proposes a model-based recursive partitioning (MBRP) algorithm to develop pedestrian exposure models. The MBRP approach combines traditional statistical methods (e.g. NB regression) with recursive data partitioning techniques commonly found in tree-based machine learning methods. This innovative approach yields a collection of exposure models stratified according to selected input variables with unique relationships between explanatory variables and exposure. The proposed method was tested on pedestrian exposure data from North Carolina significantly improved predictions of pedestrian volumes by approximately 10%. Therefore, the MBRP algorithm presents a promising tool for advancing pedestrian safety analyses in practical applications.