Research On Influence Factor Analysis And Prediction Of Freeway Traffic Accident Based On Machine Learning

In recent years,China’s freeways have developed rapidly,but the situation of traffic safety management is very serious.Frequent and serious freeway traffic accidents have become an urgent problem to be solved.At present,the wide application of machine learning algorithm provides a good idea for solving this problem.The analysis of freeway traffic accident data through machine learning model can help us find the spatial-temporal distribution characteristics and influencing factors of freeway accidents,and then predict them,so as to carry out traffic accident prevention and control more accurately and effectively through the improvement and measures of freeway design and management.Therefore,based on the data of Shaoguan section of Beijing Hong Kong Macao expressway,Leshan Guangzhou Expressway and Guangdong section of Xuguang Expressway in recent three years,this paper focuses on the analysis of influencing factors of freeway traffic accidents,the prediction of accident prone points and the real-time prediction of accidents by using machine learning algorithm.Firstly,according to the composition and operation process of expressway traffic system,this paper makes a qualitative analysis of the influencing factors of traffic accidents.On this basis,the static and dynamic characteristics that may be related to the occurrence of accidents are generated through the processing of the data of the above three expressways,and then the relationship between some characteristics and the timespace distribution of freeway traffic accident frequency is analyzed through descriptive statistics,hypothesis testing,visualization and other methods.Secondly,this paper focuses on the traffic accident prone points in expressway.Also,based on the data of the above three expressways,this paper discusses the prediction method of traffic accident prone points based on the static characteristics of expressways by constructing machine learning models,and then analyzes the relationship between several road section characteristics and whether the road section is a traffic accident prone point by using sensitivity analysis and other methods.Finally,this paper introduces dynamic characteristics and studies the real-time prediction method of freeway traffic accidents.Taking whether an accident occurs in a certain section of a road in a certain period of time and under certain environmental conditions as the prediction object,the prediction effects of various machine learning models are compared and analyzed in combination with dynamic characteristics and filtered static characteristics,and then the influence of weather conditions,time and other factors on the occurrence of freeway traffic accidents is analyzed through sensitivity analysis and other methods.27 figures,20 tables,51 references.