Research On Short-term Traffic Flow Prediction Based On Optimized Model With Improved Grey Wolf Optimizer

In recent years,with the rapid growth of China’s economy and the improvement of people’s living standards,the number of motor vehicles has been increasing day by day.There is a contradiction between the limited road resources and the increasing number of vehicles,which results in a series of complex traffic problems.In order to solve the difficult problem,Intelligent Transport System(ITS)cames into existence.With its powerful ability of comprehensive analysis and intelligent guidance ability,it can help car owners reasonably plan their travel routes,avoid traffic congestion and improve road traffic efficiency.As the basis of ITS,short-term traffic flow prediction has important research significance.Accurate traffic flow prediction can help traffic management bureau,logistics company,car owners know the traffic situation in advance,which is of great help to reasonable route planning and maintenance of social traffic order.In this context,in order to improve the accuracy of the short-term traffic flow prediction,this thesis first improves the Grey Wolf Optimizer(GWO),and uses the Improved Grey Wolf Optimizer(IGWO)to optimize the Support Vector Regression(SVR)and Elman neural network traffic flow prediction models.The proposed combined optimization model is used to predict the traffic flow of urban intersections and expressways,which verifies that the addition of IGWO can effectively improve the prediction accuracy of the optimized model.Specifically,the main research work of this thesis is as follows:(1)Aiming at the disadvantage that the linear convergence factor of GWO can not match the global and local search well in the optimization process,a nonlinear convergence factor based on tanh function is proposed to improve it.Furthermore,the gray wolf position update mechanism with weight is used to improve the convergence speed of GWO.Ten international standard test functions are used to test the optimization performance of the improved algorithm,which verifies that IGWO algorithm has better performance in convergence speed and optimization accuracy.(2)Aiming at the super parameter optimization problem of SVR model in predicting traffic flow,the proposed IGWO algorithm is used to optimize the penalty coefficient C and kernel parameter γ of SVR model,and the IGWO-SVR traffic flow prediction model is constructed.Considering the daily periodicity and neighborhood correlation of traffic flow,the daily traffic flow of urban road intersection is predicted in different intersection environments.The simulation results show that,compared with the unoptimized model,the proposed IGWO-SVR model has a significant improvement in prediction accuracy.(3)Aiming at the disadvantage that Elman neural network model is easy to fall into local optimum when predicting traffic flow,the IGWO-Elman traffic flow prediction model is constructed by using the proposed IGWO algorithm to optimize the initial weights and thresholds of Elman model.Considering the weekly periodicity,daily periodicity and neighborhood correlation of traffic flow,the multi-step prediction of expressway traffic flow is carried out.The simulation results indicate that the addition of IGWO optimization algorithm helps to reduce the prediction error of Elman model,and IGWO-Elman model still maintains a high prediction accuracy in the future long-term traffic flow prediction.