Study on predicting the severity of traffic accidents in different grades of rural highways
There are certain differences in the characteristics of traffic accident severity on different grades of rural roads.To accurately analyze the influence of various factors and their combinations on the occurrence of serious traffic accidents on rural roads,11 factors such as road type,roadside protection facilities,period,weather conditions,and accident location were taken as independent variables,and accident severity was taken as the dependent variable.First,the Apriori association algorithm with improved directional constraint was used to find out the internal relationship among the influencing factors,and the combination of key factors was solved.Then SMOTE oversampling method was used to equalize the data samples,to improve the generalization performance of the forecast model.Finally,random forest,GBDT,and XGBoost prediction models were constructed to predict the severity of rural road accidents.The results show that the operation efficiency and mining accuracy of the improved Apriori algorithm are greatly improved.For the county road,the probability of particularly serious accidents at the intersection of two roads with wet roads during the day increases to 1.86 times;For rural roads,the probability of a particularly serious accident in the early morning on the road with a speed limit of 50 km/h and no lighting increases by 2.3 times.Compared with random forest and GBDT models,the XGBoost model has the best performance in accuracy,recall,precision,and F1 score.Lighting conditions,roadside protection facilities,road types,and weather conditions are important factors affecting the severity of rural road accidents,and there are significant interaction effects among the influencing factors.The same combination of factors has different effects on the accident severity of different grades of rural roads.The probability of particularly serious accidents on rural roads is 7.3 percentage point higher than that on county roads when driving on a single lane on rainy days.
safety engineeringrural highwaysimproved Aprioriensemble learningprediction of accident severity
NETL
NSTL
万方数据
维普
