Research On Prediction Algorithm Of Freeway Short-term Traffic Flow

City freeway,the main city to the vehicle to provide fast access service in the short distance,traffic fluctuations,high degree of coupling and ordinary road,so the short-term traffic flow forecast can bring great convenience to the city traffic.Short term prediction of traffic flow is an important precondition of traffic control and traffic guidance,and it is also the main direction of this paper.Traffic flow is usually characterized by periodicity,uncertainty and high nonlinearity,and the neural network has good nonlinear mapping ability and learning ability.Therefore,this paper will focus on the neural network based short-term traffic flow prediction algorithm.In neural network,BP neural network is widely used as the core part of feed forward neural network,but it also has some shortcomings,such as local minimum and slow convergence.To this end,this paper will be aimed at these shortcomings of the BP neural network to improve,and the optimization of the algorithm for the short term traffic flow forecasting.Firstly,describes the basic concept of traffic flow forecasting,including traffic flow of three important parameters and the characteristics of traffic flow,introduces the method of data acquisition and pretreatment of traffic flow,studies a variety of freeway dynamic traffic model,and gives four kinds of index evaluation and prediction.Secondly,a detailed description of the BP neural network short-term traffic flow forecasting model building process,including input and output generation determine the node number and the number of hidden nodes and the sample database,forecasting model simulation of the actual highway traffic flow data using BP neural network to identify the good,and the prediction results were confirmed by analysis.By using BP network algorithm to Expressway short-term traffic flow prediction is feasible,but the effect remains to beimproved.Then,the wavelet neural network model using a wavelet function instead of BP neural network hidden layer Sigmoid function to continue the simulation experiment,and compared with the BP algorithm,to further improve the performance of prediction algorithm.Finally,the MMAS algorithm and wavelet neural network fusion,using ant colony algorithm to provide a set of relatively optimal power threshold for the wavelet neural network established by the simulation experiments,the results show that the wavelet neural network by ant colony algorithm can be more snugly fitting out the variation of traffic flow,there are some practical value.