Research On The Relative Importance And Spatio-temporal Heterogeneity Of Factors Influencing The Severity Of Bicycle Accident

Bicycles are considered to be a mode of transportation that conforms to the concept of sustainable development for motor vehicles.Cycling helps the transportation system to save energy and reduce emissions,and it can also ease the traffic pressure on the crowded transportation corridors in densely populated areas in cities.However,bicycles are a vulnerable group among road participants,and their potential safety hazards have become more prominent,and traffic accidents with motor vehicles have become more and more serious.As cycling has become more and more popular in many cities in the United States,related traffic safety has also attracted more and more public attention.Therefore,it is necessary to conduct in-depth research on the severity of injury to cyclists in bicycle accidents.This paper takes the city of Chicago in the United States as the research object and analyzes the data of Chicago bicycle and motor vehicle accidents from 2015 to 2020.Starting from the complexity of the accident influencing factors and the unobserved heterogeneity,the gradient boosting decision tree(GBDT)Model and time-space weighted logistic regression(GTWLR)model,revealing the relative importance and temporal and spatial heterogeneity of factors affecting the severity of cyclists’ injuries,and propose corresponding traffic safety improvement measures from an empirical point of view to prevent accidents,To reduce the severity of accidents and improve road traffic safety.Based on the research data,first select 13 potential variables that may affect the severity of cyclists’ injuries in the accident from the four aspects of people,vehicles,roads and the environment,and build a GBDT model to study the relative importance of factors affecting bicycle and motor vehicle accidents.Analyze the relationship between significant variables and the severity of injury to cyclists.The results show that there are 7 variables that are significant variables in both the Logit model and the GBDT model.The relative importance of the three variables,the type of lane,the age of the cyclist and the time of the accident,ranks in the top three.Through a comparative analysis with the Logit model,the effectiveness and reliability of the GBDT model in the analysis of the severity of traffic accidents are verified from the classification performance.Secondly,select the top 10 variables in the GBDT model in terms of relative importance as the significant variables,conduct empirical analysis on the temporal and spatial heterogeneity of the factors affecting the severity of the accident,and construct the GTWLR model,which will pass the local coefficients of the 7 significant variables of the non-stationarity test Perform spatial visualization to analyze the spatio-temporal non-stationarity that affects the severity of injury to the cyclist.Finally,based on the research results,suggestions and measures for improving bicycle traffic safety are put forward from the aspects of safety education,safety management and technology.