Critical conflict probability: A novel risk measure for quantifying intensity of crash risk at unsignalized intersections

Abstract

A significant number of traffic crashes are reported at unsignalized intersections. However, in developing countries, challenges such as underreporting and limited crash data hinder the direct correlation of traffic conflicts with reported crashes for effective safety analysis. To address this, the study introduces Critical Conflict Probability (CCP) as a novel metric to quantify the intensity of conflict risk at unsignalized intersections. Higher CCP values indicate a greater likelihood of crash risk. CCP is derived from Post-Encroachment Time (PET) using the Generalized Extreme Value (GEV)-based extreme value theory (EVT) modeling framework. The CCP values are modeled as a function of traffic flow and driving behavior variables using three approaches: fixed parameters, random intercept, and grouped random parameters Beta regression models. The results revealed grouped random parameters Beta regression model as the best fit, highlighting the importance of accounting for spatial unobserved heterogeneity. As a practical outcome, the study develops a CCP-based intersection prioritization framework to rank and identify critical intersections within a traffic network, enabling traffic planners to improve safety management in data-scarce environments.