| With the rapid expansion of freeway mileage in China,freeway traffic crashes occur frequently,and freeway traffic safety has been a major difficulty in traffic management over the years.As the acquisition of large traffic information data turns into possible,more and more researchers are strongly interested in the study of freeway real-time crash risk based on short-time traffic flow.However,in the research on freeway crash risk,Researchers rarely consider the relationship between the dynamic characteristics of traffic flow and the risk of traffic crashes under different time,ignoring the influence of time factors(working days and rest days)on freeway crash sign factors identification and real-time crash risk prediction,which is difficult to provide precise guidance for freeway safety management under different time.Based on this,this paper uses freeway traffic crash data and matching highprecision traffic flow data to research the relationship among traffic flow characteristic before traffic crashes and traffic crash risk under different time,and explore the heterogeneity of freeway crash sign factors under different time and different macros traffic flow states,and construct freeway real-time crash risk prediction model under different time and different macro traffic states.The following are the major research contents of this paper:(1)Data preprocessing of freeway traffic crash data and traffic flow and characteristics analysis of traffic crashAccording to the data requirements for real-time freeway traffic crash risk analysis and prediction,select freeway traffic crash data and traffic flow data,select and match the sample data based on the matched case-control method,then obtain the essential data used in the paper.According to the characteristics of traffic flow,the study time is divided into working days and rest days,and traffic flow is divided into unsaturated traffic flow state and saturated traffic flow state based on the six service levels,then statistical analysis methods are used to analyze the crash characteristics under different time and different traffic flow states.(2)Recognition of traffic flow variables of crash signs under different time based on random forest algorithmWith reference to the selection of traffic flow variables in existing related studies,this paper selects some traffic flow variables,such as the average and standard deviation of traffic flow rate occupancy rate and etc.,and normalizes the data.Subsequently,based on random forest algorithm,the importance of traffic flow variables is evaluated,and the variables with higher importance are selected as the traffic flow variables of crash signs.Analyzing the traffic flow variables of crash signs,the results show that there are obvious differences in the traffic flow variables of crash signs under different time and different traffic flow states,and there are differences in the mechanism of crash occurrence.(3)Construction of freeway real-time crash risk prediction model under different time based on support vector machineFreeway real-time crash risk prediction models were constructed on the data sets of working days,rest days under different traffic flow states,and the ROC curve and AUC value were used to evaluate the prediction accuracy of the model.The results reflect that the constructed model has a certain prediction accuracy and can be applied to freeway real-time crash risk prediction.At the same time,it is compared with the single model constructed by all accidents without distinguishing time and traffic flow state.The results show that the prediction accuracy of the model constructed in this paper is higher.In addition,in order to verify the rationality of screening important traffic flow variables based on the random forest algorithm in this paper,a model was built using unscreened traffic flow variables and compared with the model in this paper.It was found that the performance of the model in this paper was better,which proved its rationality. |
PDF Full Text Request