The Interactive Mechanism and Coordinative Pre-Alarming Method of Traffic Congestion and Crashes Based on Turbulence Theory

This paper employs the micro dynamic traffic flow and matches crash data of freeways within urban clusters as samples and introduces the turbulence theory into traffic field research. Firstly, the spatial and temporal trajectories of traffic congestion and crashes are described by means of data visualization methods. Secondly, the network model is applied to explore the spatiotemporal correlation and interactive trend between the congestion and crashes. Then, the Decision Tree and Random Forest classifiers were exploited to conduct the joint causation inference of traffic congestion and crashes, and to develop the correlation model between the causal factors. Finally, the coordinative inference and pre-alarm models of traffic congestion and crashes of urban cluster freeway were established. The maximum accuracy of the inferential models reached 86%, and the minimum error of the coordinative risk prediction models of traffic congestion and crashes was as low as 0.75, matching the requirements of engineering practices.