Analysis Of Influencing Factors Of Urban Traffic Accidents And Research On Driver Characteristics

With the accelerating development of urbanization in China,the number of motor vehicles continues to rise throughout the country,and cities are also facing increasingly serious traffic accidents.This paper provides systematic accident prevention and management suggestions for the relevant departments of urban traffic management in my country by analyzing the causes of urban traffic accidents and researching the characteristics of drivers.Firstly,this paper studies the influencing factors of urban traffic accidents.Based on the data of traffic accidents in a city,the analysis is carried out from five aspects including driver,time,place,vehicle and environment.The semi-parametric logistic model based on Group LASSO is used to analyze the influence mechanism of each variable in the model.Compared with the traditional logistic model and other machine learning algorithm models,it is found that the model has greatly improved the prediction effect on the basis of high interpretability.And through the analysis of association rules,the association between multiple factors is found,and a model containing interaction terms is established to explore the coupling effect of accident factors.Secondly,this paper conducts a study of the driver’s characteristics.Based on driver data,an integrated two-stage fusion classification model of gradient boosting decision tree and factorization machine is used.Compared with the single model,the model improves the accuracy and operation speed.Then,the model classification result is converted into a scorecard to judge the risk of a driver’s traffic accident.Finally,we construct individual driver portraits and driver group portraits based on the scoring results,and give the characteristics of the drivers.The innovation of this paper is that many unconsidered indicators are introduced,the prediction ability is improved by semi-parametric model,and the interaction of influencing factors is considered.And the driver risk scoring card is innovatively proposed to build a driver portrait..The research results of this paper can not only provide effective suggestions for individual drivers and related traffic management departments,but also have reference significance for the modeling of traffic accidents and related research on drivers.