An Enhanced Empirical Bayesian Method for Identifying Road Hotspots and Predicting Number of Crashes
The Empirical Bayesian (EB) method has been widely used for traffic safety analysis. It is well known that the EB method is powerful in handling the regression-to-the-mean bias that would often arise in traffic safety analysis. A prerequisite for applying the EB method for the estimation of the safety of a road segment is to identify a group of similar road segments. In this article, the authors intend to enhance the EB method by incorporating a similarity measure based on the Proportion Discordance Ratio (PDR) into the procedure to identify similar road segments safety wise. Specifically, a methodology to assess and objectively quantify similarity among road segments based on crash patterns is developed, where each crash pattern contains a unique combination of selected crash-related features. Improvement in predicting the number of crashes that would occur in road segments by applying the EB method enhanced by the PDR is demonstrated through a case study.
- Record URL:
-
Availability:
- Find a library where document is available. Order URL: http://worldcat.org/issn/19439962
-
Supplemental Notes:
- © 2018 Taylor & Francis Group, LLC and The University of Tennessee. Abstract reprinted with permission of Taylor & Francis.
-
Authors:
- Lee, Alexander S
- Lin, Wei-Hua
- Gill, Gurdiljot Singh
- Cheng, Wen
- Publication Date: 2019-9
Language
- English
Media Info
- Media Type: Web
- Features: References;
- Pagination: pp 562-578
-
Serial:
- Journal of Transportation Safety & Security
- Volume: 11
- Issue Number: 5
- Publisher: Taylor & Francis
- ISSN: 1943-9962
- EISSN: 1943-9970
- Serial URL: http://www.tandfonline.com/loi/utss20#.U2qTJPldWCm
Subject/Index Terms
- TRT Terms: Bayes' theorem; Crash characteristics; Geometric segments; High risk locations; Mathematical prediction; Risk assessment; Safety analysis; Traffic safety
- Subject Areas: Highways; Safety and Human Factors;
Filing Info
- Accession Number: 01715113
- Record Type: Publication
- Files: TRIS
- Created Date: Aug 28 2019 5:16PM