Work Zone Crash Cost Prediction with a Least Median Squares Linear Regression Model

Abstract

Traffic safety considerations are an important component of work zone design and implementation plans. This study aimed to develop a safety estimation approach on the basis of work zone scheduling and configuration parameters to help engineers and decision makers during the work zone traffic management design stage. This approach has its basis in the fusion of multiple years of statewide work zone records, work zone–related crash reports, and traffic detection data. First, a matching algorithm was used to integrate those data sets to provide detailed circumstances of work zone crashes. Second, the actual vehicle miles traveled in work zones were calculated and used as the exposure measure. The work zone crash costs also were calculated. Third, a least median squares linear regression model was used to build the crash cost prediction tool on the basis of the work zone configurations and vehicle miles traveled. This model was tested, and the results indicated that it could accurately predict work zone crash costs. Recommendations for work zone scheduling and configuration parameters therefore could be made. Although the Wisconsin Lane Closure System work zone configurations are used for operational purposes and differ from the ones in the Manual on Uniform Traffic Control Devices, other types of work zone configurations in similar systems can be incorporated easily into this approach. Thus this approach would be easy to adopt for use with the existing traffic management plan process as a proactive method to avoid high-risk work zone scenarios.