Severity analysis of crashes on express lane facilities using support vector machine model trained by firefly algorithm

Abstract

Objective Express lanes (ELs) provide an alternative way for improving the capacity of the existing freeway network without considerably expanding the roadway footprint. Although much research has been done to explore factors contributing to crashes on these facilities, there is not much discussion on factors influencing their injury severity. This study explored factors influencing the injury severity of crashes on EL facilities. Method A Support Vector Machine (SVM) model trained by the Firefly Algorithm was used to identify factors influencing the injury severity of crashes on EL facilities. The analysis was based on three years of crash data (2012–2014) from four EL facilities in California, totaling 61 miles. Results The results indicated that the following factors increased the probability of an injury or a fatality: concrete barriers, high average annual daily traffic, rolling or mountainous terrain, weekend, adverse road surface condition, and nighttime condition. Moreover, wide right and left shoulder widths decreased the probability of having an injury or a fatality. Conclusions The results provide insights into the influence of different geometric characteristics and crash-related factors on the severity of crashes on EL facilities. The study findings may assist agencies to better understand the impacts of factors contributing to injury and fatal crashes on EL facilities and implement strategies to reduce the severity of these crashes.