This paper collected incident data from Ning-Tong (NT) freeway in Jiangsu province of China, and divided them into two categories: one was traffic accident data, the other was vehicle assistance data. Distribution estimation results indicated that traffic accident data obeyed log-normal distribution and log-logistic distribution, and vehicle assistance data obeyed logistic distribution after a Box-Cox transformation. A linear model was then constructed for traffic accident data by a stepwise regression approach. The obtained linear model could provide a preliminary prediction of incident duration when limited information about the incident is known. A methodology using historical incident database was then proposed to give updated predictions as the incident lasts. Two models, the linear one and the proposed one, were compared by case studies. It reveals that the former one provides rough and static predictions while the latter one gives more precise and dynamic predictions.
A Methodology for Freeway Incident Duration Prediction Using Computerized Historical Database
The Twelfth COTA International Conference of Transportation Professionals ; 2012 ; Beijing, China
CICTP 2012 ; 3463-3474
2012-07-23
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Predicting Freeway Incident Duration Using Machine Learning
Springer Verlag | 2020
|Hazard based models for freeway traffic incident duration
Online Contents | 2013
|Hazard based models for freeway traffic incident duration
Elsevier | 2012
|