Exploring impact patterns of freeway incidents: estimation time delay using Tobit model, machine learning, and deep learning framework

抄録

To maintain freeway traffic, it is critical to model and attain better understanding of the non-recurrent congestion caused by freeway incidents. This research develops prediction models to estimate the impact of incidents on freeway traffic in terms of impact duration. The model can enable the provision of more accurate traffic information to drivers and management agencies according to the estimation of queue lengths and associated delays. Based on the vehicle detector data derived from the Freeway Bureau and the records of emergency freeway incident responses during 2015 and 2020, this study seeks to explore the impact patterns of incidents on freeway traffic. The Tobit model, machine learning, and a deep-learning framework are adopted to estimate the delay caused by freeway incidents. The prediction of impact duration can contribute to several aspects of traffic management, such as improving the dissemination of traffic information to mitigate traffic congestion.