Prediction Of Highway Tunnel Accident Severity Based On Improved PSO-BP Neural Network

Traffic safety has always been a key concern in the transportation industry,because traffic accident s,especially accident s with fatal injuries,would cause huge losses in manpower,economy and society and numerous research has been conducted to improve traffic safety.Among the field of traffic safety research,the analysis of traffic accident severity is a key part because it reveals the relationship between the accident severity and its various explanatory variables.As a highway accident-prone section,tunnel internal rescue is more difficult than open roads section,and as a result,severe accidents have occurred frequently in tunnel.Therefore,it is of great significance to continue the in-depth study of tunnel traffic accidents,to explore the relationship between various factors and the accident severity,to improve the safety of traffic within the tunnel.This thesis analyzes the tunnel accidents occurred in Norwegian the Norwegian Open Data Center(NVDB)and establishes a tunnel accident severity forecasting index system based on the results of the feature analysis.Ten indicators such as tunnel length and annual average daily traffic volume are selected as the input variable of the predictive model.The tunnel accident severity level is used as an output variable for the predictive model.A prediction model of tunnel traffic accident severity based on improved PSO-BP neural network is constructed.With the help of MATLAB software,the screened sample data is trained and tested,and the improved model is compared with the model prediction before improvement.It is found that the optimized model has better performance and higher accuracy of the prediction than the traditional standard BP model.