Study Of Short-term Traffic Flow Forecasting Model Based On Neural Network

With the development of Chinese urban urbanization process and the growth of the automobile industry, the number of cars in China showed a trend of rapid growth. Too many motor vehicles has brought a series of problems:traffic jam, frequent traffic accidents, low energy utilization and exhaust gas pollution, etc. Among them, the traffic congestion problem is particularly serious. At present this problem still did not get a better improvement. Traffic flow forecasting is the critical step in the process of traffic management and control, accurate prediction of short-term traffic flow data are of great practical significance.Traffic flow is a complicated multi-variable and time-varying nonlinear parameter. Now there are many forecasting models which can realize traffic flow forecast. The biggest problem of a single prediction model is that the accuracy is not high. In view of the city traffic congestion problem, this paper analyzed the traffic flow data. Based on characteristics of time and space, this paper put forward the traffic flow prediction model based on neural network, and took the actual traffic flow data of the a crossroads in Guangzhou as a sample. Using the BP neural network for training and simulation, the conclusion showed that the prediction model can get a good accuracy.This paper took Guangzhou city road traffic flow data as sample, the three different kinds of MATLAB simulations are showed:1. Using the traditional BP neural network model for the simulation, and different number of hidden layer neurons number error precision were used to show the accuracy of the model.2. Using the improved BP neural network model for simulation, and the adaptive variable step size algorithm is used to improve the slow convergence of BP neural network.3. Using the BP neural network model based on genetic algorithm for the simulation. Because of the global optimization characteristic of the genetic algorithm, this model can search for the optimal BP neural network model, which can achieve high prediction accuracy and shorter convergence time prediction effect is achieved.