01590266

Component

Zero-inflated Negative Binomial models with site-specific random effects are developed with the panel data to analyze hourly crash frequency on highway segments with one-mile length in average. In addition to the refined scales in temporal domain, the new crash frequency prediction models can also consider unbalanced panel-data structure and random nature of the observations. The real-time traffic, weather and road surface condition data, primarily from the Road Weather Information System (RWIS), are incorporated along with the road characteristics. The RWIS data is available on most major highways around the United States as well as some other countries. Therefore, this study introduces promising engineering technology to develop refined-scale crash frequency models, which can be conveniently applied to many other major highways. The proposed models are demonstrated by studying one portion of Interstate I-25 in Colorado, from which some interesting insights for crash frequency prediction are made available, adding to the understanding of crash risks on major highways.

This paper was sponsored by TRB committee ANB20 Standing Committee on Safety Data, Analysis and Evaluation.

01584066

16-3729

English

17p

2016

Transportation Research Board 95th Annual Meeting

Web

References (42) ; Tables

Crash rates; Highway engineering; Highway safety; Mathematical prediction; Real time information; Road weather information systems; Weather

Colorado

Highways; Safety and Human Factors; I81: Accident Statistics; I82: Accidents and Transport Infrastructure

Transportation Research Board Annual Meeting 2016 Paper #16-3729

TRIS, TRB, ATRI

Jan 12 2016 5:37PM