A comparative analysis of black spot identification methods and road accident segmentation methods

Indicating road safety-related aspects in the phase of planning and operating is always a challenging task for experts. The success of any method applied in identifying a high-risk location or black spot (BS) on the road should depend fundamentally on how data is organized into specific homogeneous segments. The appropriate combination of black spot identification (BSID) method and segmentation method contributes significantly to the reduction in false positive (a site involved in safety investigation while it is not needed) and false negative (not involving a site in safety investigation while it is needed) cases in identifying BS segments. The purpose of this research is to study and compare the effect of methodological diversity of road network segmentation on the performance of different BSID methods. To do this, four commonly applied BS methods (empirical Bayesian (EB), excess EB, accident frequency, and accident ratio) have been evaluated against four different segmentation methods (spatial clustering, constant length, constant traffic volume, and the standard Highway Safety Manual segmentation method). Two evaluations have been used to compare the performance of the methods. The approach first evaluates the segmentation methods based on the accuracy of the developed safety performance function (SPF). The second evaluation applies consistency tests to compare the joint performances of the BS methods and segmentation methods. In conclusion, BSID methods showed a significant change in their performance depending on the different segmentation method applied. In general, the EB method has surpassed the other BSID methods in case of all segmentation approaches.