Identifying Leading Indicators of Bus Collision Rates through Temporal and Spatial Analysis of Operator Behaviors Within the Washington Metropolitan Area Transit Authority Compact Area

Abstract

This study aims to establish leading indicators for the bus collision rate by analyzing operator behaviors and their relationship to collision incidents within the study area. Using two comprehensive datasets—the National Transit Database (NTD) reportable bus collision dataset and the bus video dataset—this research investigates temporal and spatial correlations between bus operator behaviors and bus collisions. The Auto-Regressive Integrated Moving Average with Exogenous Variables (ARIMAX) model is employed for temporal analysis to identify the relationship between bus video events and the collision rate, while spatial analysis uses the Geographically Weighted Regression (GWR) model to reveal localized patterns of risky bus operator behaviors, and their correlation with bus collisions, with the built environment considered. The ARIMAX model identifies the number of bus video events as a leading indicator of the bus collision rate. The GWR model identifies that certain bus operator behaviors such as late response to risky situations, incomplete stop at a stop sign, and eating food or drinking while the vehicle was in motion are significantly associated with the bus collision rate in certain areas, while pedestrian and cyclist interactions show high localized positive coefficients over the widest area. These findings underscore the importance of targeted safety interventions and continuous driver training to enhance transit safety. By integrating advanced statistical models, this study provides actionable insights for developing proactive measures to mitigate bus collisions and improve overall transit safety.