Study On The Influencing Factors And Improvement Measures Of Expressway Traffic Accident Severity

The construction of freeway in our country is developing rapidly.While freeway bring convenience to travel,they also cause many traffic safety problems.Active traffic safety management and control is a current research hotspot in the field of traffic safety.It has a major practical effect on effectively preventing traffic accidents and reducing the loss of life and property caused by accidents.However,the severity of traffic accidents is the result of the comprehensive influence and action of many factors such as people,vehicles,roads,and the environment.Therefore,the establishment of a freeway accident severity prediction model has important theoretical and practical significance.In order to study the severity of freeway accidents and dig deeper into its influencing factors,first of all,this paper summarized the elements of the freeway traffic system,traffic flow characteristics,and traffic accident prevention methods.Based on related theories such as traffic engineering,traffic safety engineering and traffic management,the existing research was analyzed from three aspects: freeway accident severity,accident severity influencing factors and accident severity prediction model.Through the collection of accident data,road design engineering drawings and video data of a certain section of the national G30 freeway in Henan Province,the potential influencing factors of the severity of the accident were organized and screened,and vehicles,people,and other vehicles illegally on the road in the original data were eliminated,and carry out statistical analysis and standardized treatment of potential influencing factors.Secondly,comprehensively considering factors such as drivers,vehicles,roads,and the environment,the Logit model commonly used in traffic accident severity prediction was selected,and the types of accidents were divided into single-vehicle accidents and multi-vehicle accidents,and multiple Logit models and hybrid Logit models were established respectively.The model predicts and analyzes the severity of the accident,and selects a better model through goodness of fit test and prediction accuracy comparison.based on the analysis of traditional accident severity models,based on random forest and gradient hoist algorithms in machine learning,a freeway accident severity prediction model was constructed,and the importance of influencing factors was ranked,and the model prediction accuracy test was passed.Chose the more suitable machine learning algorithms.Finally,the optimal model in the traditional accident severity prediction model and the machine learning-based accident severity prediction model was compared and analyzed through the K-fold cross-check method to determine the final severity of the freeway accident prediction model based on the gradient hoist.And the backward selection method was used to screen and analyze the core influencing factors of the severity of accidents,combined with traffic system elements such as people,vehicles,roads,and the environment,and proposed targeted improvement measures from traffic engineering and traffic management in order to reduce accidents.Provided a theoretical basis for reducing the casualties and property losses caused by the accident.