Research On Influencing Factors Of Traffic Accidents Based On Neural Network Algorithm

With the development of economy,the road infrastructure cannot meet the growing demand for vehicle traffic.Road traffic accidents are frequent which threats people’s safety.How to reasonably predict the severity of traffic accidents and avoid the occurrence of traffic accidents is the top priority of road traffic safety research.This thesis selects appropriate people,vehicles,roads,and environmental factors,and uses Back Propagation(BP)neural network,random forest,BP neural network combined with Genetic Algorithm(Back Propagation With Genetic Algorithm,BP-GA)to predict the severity of traffic accidents.Then,this thesis analyzes the most significant influencing factors of traffic accidents through sensitivity analysis.The main work is as follows:(1)This thesis uses the methods of literature summary and SPSS correlation analysis to select appropriate factors as research variables.Then classifies and predicts minor accidents,serious injury accidents,and fatal accidents.(2)Predicts the severity of traffic accidents by using BP neural network,random forest and BP-GA neural network respectively.In particular,the BP-GA neural network has been deeply studied.Combines the advantages of the genetic algorithm to search for the optimal solution and the advantages of the nonlinear and strong mapping ability of the BP neural network and uses the genetic algorithm to update the weights and biases of the BP neural network.The BP-GA neural network model of the combined algorithm is constructed to predict the severity of traffic accidents..(3)Uses the sensitivity to calculate the sensitivity value of each input variable of the BP-GA neural network,and determines the most significant influencing factors affecting the severity of traffic accidents through the sensitivity value.Then it uses sensitivity calculation method and mathematical statistics method to analyze how these significant influencing factors affect the severity of traffic accidents.Finally,this thesis gives the prediction strategy to prevent traffic accidents.The research results show that the BP-GA neural network model has the best prediction effect,with the highest accuracy rate of about 85%,the BP neural network prediction accuracy rate of about 83%,and the random forest prediction accuracy rate of about 66%.The most significant road and environmental factors that affect the severity of traffic accidents are speed limits,lighting and road type.This thesis enriches the method system of intelligent identification and prediction of traffic accidents,combines BP neural network and genetic algorithm,and obtains important factors and corresponding laws that affect the severity of traffic accidents,which provides technical reference for microscopic research on traffic safety.The provided BP-GA prediction model can help predict the severity of accidents,and the occurrence and severity of accidents can be reduced through early warning and preventive strategies.