After an accident occurs, traffic risk propagates upstream along the roadway. Within the spatiotemporal impact area of the accident, traffic flow accident risk experiences fluctuations with uncertainty. Accurately describing the process of traffic accident risk propagation on the road and quantitatively analyzing it will provide a solid theoretical basis for traffic risk management measures. Therefore, this study focuses on the risk propagation mechanism and prediction methods after traffic accidents in highway diversion areas. Firstly, this paper introduces the concept and indicators of risk propagation. Secondly, through simulation experiments, the driving scenarios in the diversion area are simulated with different traffic volumes, and the regularities of risk propagation are studied by observing changes in risk indicators. Finally, using a simulated dataset as input, a GRU (Gated Recurrent Unit)-based accident risk prediction model is constructed. Experimental results indicate that in predicting accident risk, GRU exhibits higher prediction accuracy compared to LSTM (Long Short-Term Memory).
|