Traffic Accident Analysis And Prediction Research Based On Improved Bayesian Network

With the rapid development of the world economy and the acceleration of the mechanization process,traffic accidents are becoming more and more frequent.In order to accurately find out the causes of traffic accidents and predict the severity of traffic accidents,this paper focuses on the current analysis of the causes of traffic accidents and the prediction of their severity.In order to solve the existing problems,research on key technologies related to traffic accident analysis and prediction based on the improved Bayesian network was carried out.The specific research contents are as follows:(1)This paper proposes a traffic accident analysis and prediction method based on the improved Bayesian network.The main ideas of this method are as follows:firstly,in view of the fact that the K2 algorithm in the traditional Bayesian network model needs manual experience to determine the order of nodes,which leads to the problem that the model is too subjective,this method sorts the factors obtained when traffic accidents occur by Gini index,The accuracy and objectivity of the model are improved;secondly,for the problem that the traditional model is not accurate enough due to lack of real-time performance,this method combines the Bayesian network with the continuous learning model constructed in this paper.When the amount of new data reaches a threshold,the method recalculates the parameters and updates the Bayesian network model,so that the model remains accurate as the data changes;and other factors to analyze and predict the severity of traffic accidents.(2)This paper uses the traffic accident dataset to test and verify the proposed method,and compares it with traditional models such as multinomial logistic regression,random forest,and Bayesian network.The experimental results show that the average absolute error of parameter learning for the Bayesian network traffic accident analysis and prediction method based on random forest proposed for the first problem is 0.0373,the average relative error is 0.0359,and the prediction accuracy rate is 78.1%.Secondly,the average absolute error of parameter learning of the traffic accident analysis and prediction method based on Bayesian network continuous learning proposed for the second question is 0.0371,the average relative error is0.0357,and the prediction accuracy rate is 79.5%.Finally,the traffic accident analysis and prediction method based on the improved Bayesian network proposed in this paper for the two problems has the best performance in terms of technical indicators.The average absolute error of parameter learning is 0.0368,the average relative error is 0.0354,and the accuracy rate reaches 80.7%..Overall,the method proposed in this paper performs better than the traditional method and the other two methods,and can analyze the cause of traffic accidents and predict the severity with high accuracy and low error.(3)This paper designs and implements a traffic accident early warning prototype system,and implements it using Java and other tools.The system has the functions of real-time traffic accident cause analysis and severity prediction,and its effectiveness has been verified by instantiation tests and feasibility.