| Identifying 'sites with promise,' also known as black spots, hot spots, or high-risk locations, has received considerable attention in the literature. This is not surprising, since there is public and professional pressure to allocate safety investment resources efficiently across the transportation system and to invest in sites that will yield safety benefits for relatively modest cost. In addition, US federal legislation requires the practice of remediating high risk locations.; A review of the literature reveals a rather disparate and non-cohesive collection of papers regarding the hot spot identification (HSID) problem. To discuss the evolution of hot spot identification issues and illuminate contemporary issues, a comprehensive survey on HSID methods is first presented using five categories. Among the many unresolved issues, three of them that need to be addressed timely are identified. In response, three studies are designed and conducted.; In study 1, three new evaluation tests are developed and described, which include the Efficiency of Prediction (EP) Test, the Method Consistency (MC) Test, and the Rank Difference Sum (RDS) Test. Use of the metrics is illustrated using a set of hypothetical intersections. The newly proposed tests, along with the conventional ones, collectively offer the ability to measure various dimensions of HSID method performance.; Study 2 aims to evaluate the performances of Empirical Bayes (EB) and two traditional methods through a controlled experiment design. Various levels of confidence are explored, and false positives and false negatives are compared across methods. The results illustrate that EB technique significantly outperforms the others.; Study 3 explores the performance of jointly using the criteria of total and excess crashes. Both EB and raw estimates are employed. Two evaluation ways and four hotspots sizes are examined. The results demonstrate that the combination method appears to be superior to the other two using single criterion in identifying hazardous locations possessing both higher crash counts and potential crash reduction.; Although the research studies here reflect an improved understanding of how various HSID methods perform and how to contrast HSID methods, further work is still needed and suggested. |
