Construction And Empirical Study Of Probit Prediction Model For Pedestrian Accidents Severity

With the continuous increase of highway mileage in China,the number of car ownership continues to increase,and traffic accidents have been in a high position.In China,most roads are modes of mixed traffic,and the traffic safety of pedestrians on the road as a vulnerable group must be particularly noticeable.According to relevant statistics,the number of traffic accidents involving pedestrians,two-wheelers,electric vehicles and other vulnerable groups accounted for more than 50% of the total number of accidents,of which pedestrian accidents accounted for 20% of the total.Then it is necessary to study the mechanism of pedestrian accidents and the factors affecting the severity of accidents,and to propose accident prevention measures and reduce the degree of pedestrian damage according to the research results.This paper takes pedestrian accidents as the research object.According to the mechanism of pedestrian accidents,it firstly analyzes the four aspects of people,vehicles,roads and environment to explore the potential influencing factors of pedestrian accidents.Then,based on the availability of data,300 people-car traffic accident data samples were extracted from CIDAS database,15 accident feature parameters were selected,the accident feature parameters were classified,and the sample data was analyzed from macroscopic and microscopic aspects.The distribution characteristics of the characteristic parameters of pedestrian accidents accidents are obtained.In order to further study the characteristics of pedestrian accidents data,this paper first analyzes the Kendall ?s tau-b correlation coefficient of the accident characteristic parameters,and determines the pedestrian age,pedestrian pace,vehicle speed,vehicle bumper height,road driving environment,Light,etc.,are explanatory variables that affect the severity of an accident.Because the severity of traffic accidents is a discrete variable,and the objective reasons for data recording,the traffic accident data generally has heteroscedasticity.Therefore,the ordered Probit model and the heteroscedastic ordered Probit model for predicting the severity of pedestrian accidents are established respectively.From the aspects of parameter estimation,goodness of fit and accuracy of prediction,the two models are compared in the severity prediction of pedestrian accidents,the heteroscedastic ordered Probit model has a wider scope of application,can handle the heteroscedasticity of data,and can also mine potential variables that have an impact on pedestrian accidents.Combining the analysis results of the two models,we can see that although the two models are different,the interpretation effect of the dependent variables is consistent when they have the same explanatory variables.The final regression model shows that in the pedestrian accidents,other variables are unchanged.When the pedestrian speed increases within 4km/h,the vehicle speed increases within 30km/h,and the bumper height increases within 0.5m.The degree of pedestrian injury is reduced.When the pedestrian age is over 41 years old and the speed exceeds 50km/h,the pedestrian injury rate increases significantly.The vehicle speed is over 60km/h.After the age of 65,the pedestrian mortality rate increases significantly.Based on the results of the above empirical research,the paper puts forward practical suggestions from the aspects of people,vehicles,roads and traffic management to reduce the accident rate and the severity of the accident.