Development and Application of Highway Traffic Congestion and Accident Risk Early Warning Model

In order to change the passive monitoring mode of the current road network operation, a portable and reusable traffic congestion and traffic accident risk pre-prediction and active early warning model was studied. After comprehensive research and study of domestic and international traffic operation state early warning algorithms, a dynamic Bayesian network was selected to build the early warning model. Firstly, nearly 4 million traffic control station and vehicle detector detection data from May to December 2019 were integrated, and data integrity test and multi-source data association were carried out to mine the risk characteristics of traffic congestion and traffic accidents on the highway network, which were used as sample data to train the dynamic Bayesian network. Through testing and analysis, the model can realize short-term traffic congestion and traffic accident risk early warning, and the early warning accuracy rate exceeds 85%. The developed model shares the early warning results with the existing business system in real time through the data interface to improve the scope of application of the model.