Transferability Study of Urban Arterial Safety Performance Functions Between Shanghai and Guangzhou

Shanghai, China, has developed a series of safety performance functions (SPFs) to analyze crash contributing factors and identify hazardous locations for the purpose of safety improvements. However, many other cities in China, such as Guangzhou have not developed local models due to lack of reliable safety data. This paper focuses on investigating the transferability of SPFs among similar cities, so jurisdictions without SPFs can more quickly conduct safety analyses. To this end, data on urban arterials in Shanghai and Guangzhou were collected, including crash data, geometric design features and traffic characteristics. Negative-binomial-based SPFs were developed separately for the cities during peak and off-peak hours. Then, model estimation results were transferred from one city to the other for crash prediction, and the prediction performance was evaluated. Results showed that local models could yield higher prediction accuracy than the transferred models. The results of likelihood ratio tests, conducted to evaluate model transferability between Guangzhou and Shanghai, suggested that the models could not be transferred directly. In order to improve transferability, the models were multiplied by calibration factors. The peak-hour models became transferable, but the off-peak models were still not transferable. To solve this problem, pooled data, composed of all Shanghai samples and various proportions of Guangzhou samples, were used to develop SPFs for transfer to Guangzhou. When over 50% Guangzhou samples were used in the pooled data, the models during both peak and off-peak hours became transferable. Findings from this paper prove that SPF models have the potentiality and possibility to transfer to other similar cities when appropriate methods are adopted to improve transferability.

• Supplemental Notes:
  • This paper was sponsored by TRB committee ANB25 Standing Committee on Highway Safety Performance.
• Authors:
  • Wang, Xuesong
  • Pei, Saijun
  • Li, Jia
  • Xie, Kun
• Conference:
  • Transportation Research Board 97th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2018-1-7 to 2018-1-11
• Date: 2018

Language

• English

Media Info

• Media Type: Digital/other
• Features: References; Tables;
• Pagination: 7p

Subject/Index Terms

• TRT Terms: Arterial highways; Off peak periods; Peak periods
• Uncontrolled Terms: Crash prediction models; Local data; Safety performance functions; Transferability of models
• Geographic Terms: Guangzhou (China); Shanghai (China)
• Subject Areas: Highways; Safety and Human Factors;

Filing Info

• Accession Number: 01664209
• Record Type: Publication
• Report/Paper Numbers: 18-02939
• Files: TRIS, TRB, ATRI
• Created Date: Mar 26 2018 2:31PM