Identifying Collision Parameters in Highway Work Zones Collisions Using Classification and Regression Trees

Highway safety modeling represents a means to study interventions that can potentially reduce the thousands of collisions that result in injuries and fatalities each year. Statistical modeling of roadways can identify the roadways where safety interventions are most needed or where the largest effects of such interventions could be felt. Work zone areas are necessary to repair and build new roads that, unfortunately, can increase the risk of collision. Prior to a work zone becoming active, statistical models can estimate the numbers of collisions that will occur as a result of these work zones. Likewise the effectiveness of safety interventions prior to deployment can be used to estimate the reduction in collisions as a result of the interventions.;The purpose of this research was to develop a method to estimate the numbers and types of collisions that occur on California roadways. This was achieved by using the classification and regression tree (CART) modeling method to identify key collision factors that result in injuries and fatalities by utilizing data from highway databases. After the factors had been identified, an empirical Bayesian (EB) model was then used to find the number of collisions that would occur on a given road. The CART method provides a weighting factor that can be used to find the number of each type of collision that will occur. Certain types of collisions are more likely to result in a serious injury or fatality can be identified though this method. Cost-benefit analysis could then be used to determine where targeted deployment of California Highway Patrolmen (COZEEP/MAZEEP) would be most effective in reducing these types of collisions.;Based on the analysis of the combined CART and EB methods, the types of collisions that are more likely to cause injury/fatality on California highways were identified as well as factors that influence collisions. The methodology developed represents a significant step in highway analysis as it combined information from several established highway databases while also utilizing several highway safety research methods. More accurate prediction models can be used to evaluate where safety interventions, such as COZEEP/MAZEEP, would have the largest impact prior to implementation. These collision models can be used to find the more optimal locations to deploy said safety interventions. Three of four highways identified types of collisions that lead to injury/fatality collisions. Only one highway, I-680, identified the underlying primary factors that are likely to lead to injury/fatality collisions. Cost-benefit analysis was used to calculate the effective cost of deployment of officers based on the numbers and types of collisions reduced. This research provides the framework that can be applied to other roadways within California.