Highway Traffic Accident Severity Prediction Based On Ordered Logit And Multinomial Logit Model

The expressway has become an important part of the transportation system in China due to its outstanding advantages such as high speed and large traffic capacity.However,with the continuous growth of the mileage and traffic volume of expressway,the traffic safety problem of highways is becoming increasingly prominent.Therefore,the in-depth analysis of highway traffic accidents is of great significance to improve the safety of highway traffic,prevent and reduce the occurrence of traffic accidents.First of all,this paper systematically analyzes the influencing factors of expressway traffic accidents from four aspects: people,vehicles,roads and environment.Secondly,based on 316 traffic accidents(K60+000 ~ K90+000)of Beijing Zhuhai Expressway(K60+000 ~K90+000),the traffic accident data and road linear data are analyzed and studied in detail,and 13 main factors are selected from road alignment,vehicle factors,traffic and environmental items.For independent variables,we use the Pearson correlation coefficient analysis method in SPSS to analyze the variables,and then divide the severity of the accident as the dependent variable into three according to the classification standard of the severity level of traffic accident and the three aspects of the accident death number,the number of injuries and the loss of direct property.Grade: direct property loss accident,injury accident and death accident.The hybrid stepwise selection method is used to screen the independent variables,and then the road traffic accident severity prediction model based on the ordered Logit model and multiple Logit model is constructed respectively,and the model is tested in proportion and the goodness of fit is tested to verify the validity of the model.Finally,the feasibility of the ordered Logit and multiple Logit models is verified by the road accident data collected.The accuracy of the ordered Logit and multiple Logit models is compared and analyzed,and the improvement measures and preventive measures are put forward by using the main influencing factors of the model analysis of the severity of traffic accidents.Through the ordered Logit prediction model,it is concluded that the location of the accident,the number of vehicles involved in the accident,the weather,the special section,the slope,the line of the road,the date of the accident are significantly related to the severity ofthe traffic accident,and the multiple ordered Logit prediction model draws a conclusion: the special section,the weather,the place of the accident,the line shape of the road.The accident date,slope and vehicle type are significantly related to the severity of traffic accidents.Through the case study,the order Logit model has the best effect on the prediction of only property loss accidents,reaching 86.52%.Many Logit models predict the best effect of the death accident,reaching 88.68%.The average prediction accuracy of multiple Logit models is better than the ordered Logit model.