Analysis And Prediction Of The Severity Of Road Traffic Accidents Considering The Influence Of Weather And Road Conditions

In recent years,the booming road traffic industry has also brought many traffic accident safety hazards.Many studies have only analyzed highway accident safety in detail but lack universal research suitable for all-road traffic accident safety.In this paper,to address the problems of difficult access to domestic traffic accident data and incomplete information coverage,we use the traffic accident data provided by the U.S.traffic system collection platform,take two dimensions of weather conditions and road conditions as the entry point,analyze the key factors influencing the severity of traffic accidents with the help of the current popular machine learning algorithms,and provide some reference for China’s road safety management.The main work and research results of this paper are as follows.The main work and research results of this paper are as follows.(1)Analysis of road traffic accident characteristics and influencing factors.The main analysis is the spatial and temporal distribution characteristics of accidents and the distribution pattern of traffic accidents under the influence of weather and environment.The association rule Apriori algorithm is used to correlate the influencing factors of traffic accidents.Through the analysis of the combination of factors acting together to influence the mechanism,the weather,air pressure,visibility,humidity and traffic signals are related in the binomial association rule;visibility and humidity are more correlated,and if a traffic accident occurs in cold weather conditions,it is likely to occur in areas with higher air pressure.(2)Comparative analysis of motor vehicle traffic accident severity prediction based on machine learning methods.After balancing the traffic accident data with data using SMOTE and Random Under Sampler algorithms,Random Forest,Decision Tree,Logistic regression model,Support Vector Machine(SVM)and K-Nearest Neighbor(KNN)classification models were built,and the results of several machine learning models were compared,and the results showed that the random forest model performed best overall.(3)Improved random forest algorithm modeling based on K-means clustering.To solve the heterogeneity problem of accident data,K-means clustering is used to process traffic accident data to obtain homogeneous data to improve the prediction accuracy.The results show that the improved random forest model,the model performs well in terms of recall and accuracy,making the imbalanced data prediction results more accurate.Finally,SHAP interpretation method is introduced to visualize and qualitatively analyze the importance of model features,and it is concluded that the main influencing factors are traffic jam distance,accident duration,and relative humidity in order of contribution.