Analysis Of The Influence Of Road Driving Environment On Severity Of Traffic Accidents

Effectively preventing serious accidents and reducing the loss of people’s lives and property in accidents are the core objectives of traffic safety research.The road driving environment is an important type of the factors affecting the pattern,frequency and severity of traffic accidents.In this paper,the relationships between road traffic environment,such as weather,road lighting conditions,road protection facilities,visibility,and the severity of traffic accidents were studied based on the accident data collected in many years in a domestic area of China.The methods of statistics,data mining,machine learning were applied to analyze the data.Firstly,the characteristics of traffic accident data were analyzed by statistical methods,and the data are preprocessed.According to the characteristics of data,several typical road driving environment scenarios were selected for subsequent modeling analysis.Secondly,the characteristics of road accident severity under different combinations of weather and light conditions were analyzed.The random parameter ordered Logit model was used to analyze and compare the influences of natural environment,road infrastructure,vehicles and other factors on the accident severity in different weather and light conditions.The results showed that human-vehicle collision and truck type always significantly affected the severity of accidents under all combinations of weather and light conditions.The probability of casualty accidents on hilly terrain,road separations and special road sections increased significantly under daytime/sunny and daytime/rainy conditions.In some parts of the road,the center and non-isolation facilities did not play a protective role,but aggravated the severity of the accidents.Thirdly,the relationships between protection facilities of traffic safety and the severity of road accidents were studied.The association rules mining algorithm optimized by hybrid particle swarm was used to analyze the causes of the accidents in different combinations of road protection facilities.The strong association rules between the factors of environment,road,vehicle and different types of accidents were identified.The results showed that with the protections of roadside and central protection facilities,the death accidents often occurred at 6am to 11 am and on hilly roads.When there are only roadside protection facilities,trucks easily caused death accidents.When only central protection facilities existed,fatal accidents often occurred on urban expressways and freeways.Fatal accidents often occurred on other types of higways where no side protection or central protection existed.According to the association rules,some suggestions were proposed to reduce the severity of accidents.Finally,the predicting method of traffic accident severity classification under different visibility condition was studied.After pre-processing and coding the accident data collected in the roads with low visibility,the XGBoost model was selected as the accident severity predicting model,as it had the best comprehensive performance after comparing its prediction accuracy with that of the logistic regression model and random forest model.The SHAP values were used to analyze the key factors affecting the accident severity in low visibility.The results showed that direct property loss,central isolation facilities,vehicles,roadside protection facilities and accident types were significantly related to the severity of accidents in low visibility.Compared with accidents occurred in the high visibility condition,the factors such as vehicle type,accident type,illegal behavior,low visibility range and traffic signals had more significant influences on the severity of accidents in low visibility.