Aggregate crash prediction model based on gravity model: Introducing crash risk distribution concept

Abstract

Crash prediction models (CPMs) can be valuable for future transportation planning decisions. This study aims to develop CPMs based on the trip distribution step of the common four-step demand models. For this purpose, the Gravity Model is used. For model calibration, the frequency of severe crashes (including the total of fatal and injury crashes) between each origin-destination (OD) pair of traffic analysis zones (TAZs) in the city of Qom in Iran has been used as the dependent variable. The number of trip distributions by purpose, traffic characteristics on the links, and road network characteristics has been used as the explanatory variables. The model validation results show a significant relationship between the mentioned variables. Therefore, in addition to predicting the crash frequency according to trip number changes in the future, the developed model in this study determines the relationship between the crash frequency with the OD characteristics of the trips that lead to crashes. This makes it possible to evaluate the impact of different travel demand management scenarios on safety so that the crash risk (i.e., crash occurrence probability) of trips distributed between TAZs is identified and prioritized and can be planned to improve or reduced them.