Identifying Hot Spots by Modeling Single-Vehicle and Multivehicle Crashes Separately

Considerable research has been conducted on the development of statistical models for predicting motor vehicle crashes on highway facilities. These models often have been used to estimate the number of crashes per unit of time for an entire highway segment or intersection without distinguishing the influence that subgroups have on crash risk. The two most important subgroups identified in the literature are single- and multi-vehicle crashes. Recently, researchers have noted that two distinct models for these two categories of crashes provide better predicting performance than models that combine both crash categories to predict crashes for an entire facility. Thus, a study was done to determine whether any difference exists in the identification of hot spots when a single model is applied instead of two distinct models. A hot spot (or black spot) is a site with an accident frequency that is significantly higher than expected at some prescribed level of significance. The data used for the comparison analysis were collected on Texas multilane undivided highways for 1997 to 2001. The study shows that modeling single- and multivehicle crashes separately predicts slightly fewer false positives and negatives than modeling them together under a single aggregated model in the hot spot identification process. Thus, it is recommended that separate models be developed for single- and multivehicle crashes for predicting crashes and for identifying hot spots.