The Application of Empirical Bayes Approach for Identifying and Ranking Hazardous Junctions Case Study:

Singapore Signalized Junctions

Abstract

The identification and ranking of hazardous road locations are important parts of road safety improvement programs. This paper describes the application of Empirical Bayes (EB) approach for identifying and ranking hazardous junctions. Accident, traffic, and junction geometric/environment data from 203 four-legged and 186 three-legged signalized junctions across western part of Singapore were collected. Accident prediction models were developed and safety of the junctions was estimated. After that, hazardous junctions were identified using probability of selecting the worst site concept and then ranked using PSI (potential for safety improvement) and LH (level of hazard) criteria. A total of 38 junctions were found as hazardous. The result shows that the use of PSI criterion is more favorable than LH criterion as it is better able to detect the top hazardous junctions with the largest number of accidents in the study period.