Study On Traffic Safety Prediction Of Mountain Expressway Based On Traffic Accident Analysis

As the most important carrier of passenger and freight transportation,the traffic safety of expressway has always been the focus of attention.In order to improve the traffic safety of expressway in mountainous areas,this paper,by digging the relationship between the expressway traffic accidents and their influencing factors,predicts the traffic safety of the expressway in mountainous areas from two aspects of accident which are the number of traffic accidents and the severity of the accidents.Based on the traffic accident data of Chongqing expressway,the prediction model of the traffic accidents and the severity of the accidents is established with the method of statistical regression,so as to reduce the traffic accident rate and the accident loss.This paper has obtained the following main results:(1)A set of road geometric data calculation method is designed based on satellite map and differential theory.According to the above method,the geometric data of 4 Chongqing expressways are obtained.Based on the principle that linear index is consistent,the accident prediction unit is divided,which verifies that the accident number of expressway prediction unit is subject to negative binomial distribution.(2)According to the statistical distribution characteristics of the number of road segment accidents,a prediction model of the number of accidents based on negative binomial regression is established.Model fitting results showed that five geometric index which are the segment length,longitudinal section length,slope length,horizontal curve radius,horizontal curve Angle have significant influence on prediction unit accidents,and the degree of influence reduced in turn.The results of likelihood ratio test and model good-of-fit test have proved that the fitting effect of Negative binomial regression model is better than that of Poisson distribution and Zero expansion negative binomial regression model.The reliability of the model is verified by comparing the relationship between the independent variables with the number of road segment accidents,which is reflected by the model and the actual data,as well as the actual and predicted accident rates of the whole section unit.The portability of the prediction model of the number of accidents is verified by the application of an example.(3)Based on the analysis results of the influencing factors of traffic accidents,10 factors were selected as independent variables from the system of drivers,vehicles,road conditions and environmental characteristics,and the accident severity prediction model based on generalized ordered logit was established according to the classification of accident levels.The results of model estimation show that weather,crash form,link type,road alignment,lighting condition,season and time have significant influence on accident severity.Based on the analysis of the advantage ratio of the model estimation results,it is found that the probability of serious traffic accidents is higher when accidents occur in bridge sections,construction sections,summer and winter.The accuracy of the model is verified by the likelihood ratio test and the comparation of the real and predicted severity levels of traffic accidents.The application of the model to an example verifies that the accident severity prediction model has universal applicability.