Aanlysis Of Road Traffic Accidents And The Influential Factors Of Severity

The traffic accidents have threatened people's lives and property,and have become an obstacle to the development of human society that should not be neglected.It is very necessary to study the occurrence mechanism of traffic accidents,especially the influential factors of the fatal and injury accidents,and propose more specific preventive measures based on these influential factors,which are of great significance for improving the level of road traffic safety management.This paper studies the influential factors of the accidents injury severity and their mutual relations from the macroscopic and microscopic perspectives,and provides some reference for traffic safety management departments to formulate policies.At the macro level,this paper use the latest panel data including 31 provinces of China from 2004 to 2015 to build panel data models.The results show that the increase of road mileage,the number of public transport vehicles for urban population,the number of street lights per kilometer and the number of medical and health facilities per 10,000 people can reduce the number and severity of traffic accidents,while the increase of per capita urban road area and illiteracy rate will increase the number and severity of traffic accidents.At the same time,for the problem of low fitting degree of panel data model of deaths caused by traffic accidents,this paper establishes a spatial lag fixed-effect panel data model.The results showed that considering the spatial correlation can effectively improve the model's fitness.At the micro level,in order to study the influence of traffic elements such as people,vehicles,roads,and environment on traffic accidents and the interrelationships among these elements,this paper adopts an ordered Logistic model and an ordered Probit model to analyse the detailed traffic accidents data.At the same time,in order to expand the ability of the regression model to interpret the dependencies between variables,the second-order interaction terms between the variables are included in the model for analysis.The results show that adding interaction items in the model can mine the dependencies between variables,and improve the model's fitness and accuracy.But it's a complex work to build a good fitting regression model,so this paper adopts the Tree Augmented Na?ve Bayesian classifier and Markov Blanket Bayesian classifier to construct the Bayesian network model,and studies the complex interrelationship between the influential factors of traffic accidents.The results show that the accuracy of the Tree Augmented Na?ve Bayesian network model is slightly higher,and the structure of the model is simple,so it takes a short time to build the model.Therefore,the model is suitable for forecasting the severity of traffic accidents.Although the structure of the Markov Blanket Bayesian network model is complicated,it can fully exploit the interrelationships among variables.Therefore,the Markov blanket Bayesian network model is more suitable for analyzing the mechanism of action between influential factors of accident injury severity.