Research On Real-Time Forecasting Model Of Highway Accident Risk Combined With Truck Factors

At present,researches on the active safety prevention and control of expressway mainly establish the accident risk prediction model based on the overall traffic flow parameters,but there is little research on the truck factor,but the truck has a great influence on traffic accidents.Therefore,it is of theoretical significance and practical value to use machine learning algorithm to establish a real-time prediction model of expressway accident risk integrating truck factors.Based on the analysis of the characteristics of freight car operation on expressway,the thesis establishes the correlation curve between freight car factors and the overall traffic flow parameters.The factors which have significant influence on the safety of expressway operation are summarized and screened out,such as the proportion of freight cars and the difference of passenger and freight speed.The I-710 freeway in the United States was selected as the research object,and the traffic accident data were classified and analyzed.Based on this,the main characteristic factors of the real-time prediction model of freeway accident risk with truck factors are determined.These variables include overall traffic flow variables,freight vehicle related variables and weather related variables.XG-Boost algorithm was used to sort the importance of all characteristic variables,and 12 of the 30 main influencing variables were found to be related to freight cars.This further confirms the importance of the van factor.Support vector machine(SVM)was used for modeling,and grid optimization method was used for parameter optimization,then the SVM classification model with different characteristic variables in different time periods was established.The performance of each classifier was compared by using confusion matrix,prediction accuracy and ROC curve.The results show that the support vector machine model based on the data from 5 to 10 minutes before the accident is better in predicting the accident.After adding the related variables of trucks,the overall prediction accuracy of the prediction model reached 89.06%,which was 11.32% higher than that without adding the related variables of trucks,and the accident prediction accuracy was 15%higher than that without adding the related variables of trucks,and the false alarm rate was 8.28% lower.In the development of highway safety early warning system,this model can be used as a module of the system to warn traffic accidents.It can also assist the highway operation safety department in the management of trucks.