Research On The Prediction Method Of Improving BP Neural Network Highway Accident Risk Level

Among the many factors affecting traffic safety,road conditions play an important role,as is the case with road traffic accident-prone sections and points.Therefore,it is important to strengthen safety audits at the road design stage in order to reduce design hazards and improve road safety.The purpose of this thesis is to explore the correlation between road accident risk and road conditions,and to study road accident risk assessment methods,which are of positive significance for enriching road traffic safety audit tools.Firstly,the connotation of road traffic accident risk is analysed,and the categories(or levels)of traffic accident risk and their classification criteria are determined according to relevant norms.Based on the analysis of the functions of the BP neural network model and its advantages,the BP neuron network intelligent classification algorithm is selected as the basis of the road accident risk assessment method,and the methodological framework of the road accident risk assessment based on BP neural network is designed.On the premise of analysing the factors affecting road traffic safety and the relationship between road conditions and traffic accidents,the method of determining the input indicators of the BP neuron network is studied,and a correlation analysis is carried out on its road traffic accident sample data,taking the highway in M province as an example,and nine road conditions that are strongly correlated with road accident risk are obtained as the input indicators of the neural network model.The thesis further investigates the construction method of the BP neural network-based road accident risk evaluation model and the optimization method of the weights and thresholds of the BP neural network based on genetic algorithm,and details the construction process of the road accident risk evaluation model with the experimental method to determine the number of neurons in the hidden layer of the neural network,the road conditions as the input and the road section risk level as the output,and the use of The model algorithm and optimisation algorithm are implemented in MATLAB.Finally,the model was trained using a large sample of highway accident data from Province M.The training effect was verified using a test set.The results show that the model has a risk rating hit rate of 84.6%.The results were compared with the traditional method Logistic model and the unoptimised BP neural network model,and the results showed that the hit rate of this model was significantly higher than that of the Logistic model and the unoptimised BP neural network model at 78.6% and 79%.