Research on Traffic Accident Severity Prediction and Analysis Based on TSO-XGBoost Optimized by QRBL
Traffic accident severity prediction is one of the research hotspots in the field of traffic safety.To clarify the crucial influencing factors of accident severity and further improve its prediction accuracy this study proposes a traffic accident severity prediction model based on Quasi-Reflex-based Learning(QRBL)combining Tuna Swarm Optimization Algorithm(TSO)and Extreme Gradient Boosting Tree(XGBoost).Firstly,the paper constructs a de-tailed dataset of 36,146 traffic accident cases by collecting and integrating traffic accident data from a city in south-east China,a city in central China,and a city on the eastern coast.Followingly,the XGBoost algorithm was used to identify the key features affecting the accident severity,and the top 20 accident features were selected as input of the accident severity prediction model according to the degree of influence of the features.Finally,the optimization per-formance of the TSO algorithm was optimized based on the QRBL strategy,so as to determine the optimal value of the hyperparameters of the XGBoost prediction model,and realize the prediction classification of traffic accident se-verity,so as to improve the accuracy and comprehensiveness of the prediction.Experimental results show that the pro-posed model is superior to other machine learning models in terms of precision,accuracy,recall and Fl score,and has significant prediction advantages.Besides,the key influencing factors are analyzed in depth through the SHAP mod-el,which provides a theoretical basis for reducing the occurrence and severity of accidents.
Traffic EngineeringAccident Severity PredictionFeature ExtractionXGBoostTSOQuasi-Reflection Based Learning
万方数据
维普
