| As one of the modern urban diseases,traffic accidents have attracted the common attention of all countries in the world.Because many traffic accidents are closely related to the human risk factors of drivers,and the driving risks of different drivers vary greatly,the risky driving behavior of drivers has become the core content of road traffic safety research in various countries,and the occurrence of traffic accidents can be reduced by identifying and predicting risky driving behavior.Based on the differences between the definition of road traffic accident given by The Traffic Management Department of Public Security in China according to the Provisions on The Handling Procedure of Road Traffic Accident and previous studies,this paper explores the relationship between traffic accidents,drivers’ characteristics,and traffic illegal behaviors from the perspectives of drivers’ age,driving age,gender,occupation,and traffic illegal behaviors.This paper aims to establish a method for road traffic managers to identify and assess drivers’ dangerous driving behavior,so that provides theoretical guidance for systematic analysis of driving safety risks.The research contents and innovations of this paper are as follows:Firstly,in view of the drivers’ attributes and characteristics of traffic illegal behaviors in summary procedure traffic accident with minor damage,a driver risk behavior evaluation model based on variational auto-coder is proposed.And the driver risk is rated by using the stochastic optimization algorithm based on second-order information.Through feature extraction and dimension reduction of the input data set,the risk of drivers in traffic accidents is divided into five grades.The characteristics of each risk level are analyzed in detail to guide for traffic management departments to determine the classification of risk levels.Secondly,based on the drivers traffic illegal behaviors in the summary procedure traffic accident in the early period,using the Survival Analysis Model based on an improved parameter optimization algorithm,this paper studied the relationship between dangerous driving behaviors and traffic accidents,to solve the problem of poor efficiency of solving algorithm in summary procedure road traffic accident survival analysis,and can realize stable convergence.Through the in-depth analysis of the summary procedure traffic accident occurrence mechanism,this paper can provide a decision-making reference for the traffic management department to carry out the control of traffic illegal behaviors.Thirdly,aiming at the characteristics of risky driving behavior of drivers with secondary responsibility in the general program road traffic accidents with relatively large damage degree,this paper proposes the topic optimization model to explore the relationship between different accident themes hidden in traffic accidents and the influence factors of illegal behaviors.In addition,the solution algorithm based on second-order information of variance reduction is proposed to achieve a better representation of the output word vector.Examples show that the algorithm has good efficiency,and it is found that lane change and reverse driving are as important risky driving behaviors as the six serious traffic violations summarized from practical experience.Fourthly,because of the problem that the practical consequences of road traffic accidents of the general program are usually serious and bring huge loss of life and property,the wavelet clustering algorithm is proposed to predict and analyze the spatial-temporal risk driving behaviors of severity level.The wavelet clustering algorithm is used to analyze and verify the data of traffic accidents and serious traffic violations in each month from 2015 to 2018 of some traffic police brigades in a city of North China,and predict the law of traffic accidents from multiple angles.The results can put forward accurate response plans for traffic management departments and improve the scientific management level. |
PDF Full Text Request