Addressing pedestrian injury risk with density peak clustering and familiar Bayesian spatiotemporal epidemiological modelling

This study tends to investigate pedestrian injury risk by combining density peak clustering (DPC) with familiar Bayesian spatial-temporal epidemiological models. The dataset was collected from 2019 to 2022 in the Chicago Metropolitan area. First, the DPC method was employed to cluster all the pedestrian crash injuries; Second, after preprocessing and buffering, a raster map and adjacent matrix of pedestrian crash injurieswere obtained; Next, the relative risk of pedestrian crash injury was considered to construct for familiar Bayesian spatial/temporal/spatial-temporal independent/spatial-temporal interaction models, and the interaction model was selected in terms of Deviance information criterion (DIC) value to explore spatial-temporal features of pedestrian crash injury. The results show that there lies in spatial heterogeneity of pedestrian crash injury risk and temporal effect varies in different years. Moreover, pedestrian injury risk is significantly concerned with weather, road surface, traffic and facility control, roadway types and driving under influence (DUI). Finally, empirical suggestions are provided to improve pedestrian safety.