Analysis Of Factors Contributing To Crash Severity On Urban Road Based On Explainable Machine Learning

Due to the improvement of the national economy,the acceleration of the process of urbanization and the progress of industrialization,the automobile industry has developed rapidly.This has also caused the road mileage and car ownership to increase year by year in my country.Corresponding traffic accidents occur frequently,which greatly affects the safety of our people’s lives has also caused immeasurable losses of national property.The inherent relationship between the various influencing factors and the severity of a traffic accident needs to be studied urgently.With the increase in the amount of traffic safety accident data and the dimension of accident data,machine learning models can improve the efficiency of data analysis and can also perform predictions on the results of the model.Visual interpretation provides relevant departments with corresponding traffic management measures to reduce the severity of accidents and reduce national property losses.Based on this,this paper proposes a method for predicting the severity of urban road traffic accidents based on interpretable machine learning.The main work is as follows:(1)Analyze the collection method of urban road accident data and the source of accident data,divide the influencing factors of the severity of the accident into four aspects: people,vehicles,roads and the environment,and analyze the influencing factors.(2)For the accident data in this article,considering the integrity of the data and the relationship between variables,11 accident influencing factors are selected as independent variables,and the severity of the accident is used as the dependent variable to preprocess the accident data.The accident data of two cars collision is selected as the research object.Analyze the correlation between the independent variable and the dependent variable of the accident data,and make a statistical analysis of the processed accident data.(3)Establish LightGBM model and random forest model,choose Accuracy,AUC and F1 as the evaluation indicators of the model.Taking into account the imbalance of accident data,the SMOTEENN algorithm is used to deal with the imbalance of accident data,and then the LightGBM model and the prediction results of the random forest model are compared with those of the logistic regression model.The results prove that the evaluation indicators of the LightGBM model and the random forest model are better than the logistic regression model,and the evaluation indicators of the LightGBM model are also better than the random forest model.(4)The LightGBM model with good predictive evaluation index results uses the SHAP model for visual interpretation,analyzes the internal relationship between various factors and the severity of the accident from the whole to the individual,and proposes targeted traffic safety improvement measures based on the analysis results.