Innovative road safety benchmarking in high-risk school vicinities using advanced data envelopment analysis

This is the authors' abstract. We don't add key points for this paper.

The proximity of schools to major roads increases road users’ vulnerability to severe traffic incidents. In environments with inadequate safety infrastructure, this risk can escalate to fatal outcomes. To address these concerns, policymakers require a structured technique to systematically compare the safety performance of road segments, identify hazardous locations with poor safety performance, and prioritize them for targeted safety interventions. The study uses a fair and novel framework to evaluate the safety performance of 54 high-risk road segments near schools in Rasht, Iran, utilizing data envelopment analysis (DEA) within the domain of road safety. Expert judgments are also integrated into DEA models using the group best-worst method (BWM) to address the limitations associated with weight flexibility. All segments are evaluated and ranked according to efficiency scores derived from a comprehensive set of indicators, including lack of safety, exposure/risk, and capacity/capability factors. Then, the causes of success and failure in achieving higher safety performance for the road segments are determined by calculating benchmark differences and applying surrogate machine-learning models based on XGBoost and CatBoost. The findings reveal that incorporating expert judgments enhances the performance of the DEA model. The numbers of fatalities, injuries, crashes, along with traffic volume are the most influential indicators, and the presence of a footbridge is the least influential in safety performance evaluation. Furthermore, the number of injuries emerged as the most common confirmed weakness among the 13 evaluation indicators across all road segments. Generally, the least efficient segments are characterized by very short distances between schools and road segments, the absence of traffic calming measures, and the lack of warning signs. A spatiotemporal ranking approach is recommended to uncover road safety deficiencies by providing deeper insight into the temporal behavior of a specific road and its users, alongside its spatial characteristics