The Research Of Short-Term Traffic Flow Forecasting Of Urban Road Based On Neural Network

Intelligent Transportation Systems(ITS) can be used in urban road network technology is one of the key short-term forecasting of urban road traffic. Real-time and accurate short-term forecasting of urban road traffic is the foundation core of urban road traffic control, vehicle induced travel and other aspects of intelligent transportation. Accurate short-term traffic forecast data, can also be applied to real-time road Intersections timing, reduce delays and increase red intersection capacity, can greatly enhance the traffic flow visualization applications timeliness, improving urban vehicle in the road network the security situation. Short-term forecasting of urban road traffic has a very broad application prospects, but the corresponding short-term traffic volume is predicted to have a random and nonlinear characteristics, etc., to make it more difficult to build a mathematical model. After careful study of the short-term prediction of the traffic characteristics, we believe that the many advantages of artificial neural networks can be its randomness and nonlinear characteristics fit. After the combination of the two theories, the establishment of urban road traffic forecasting model based on neural network short, and the model is validated instance, the specific contents are as follows:First, the amount of short-term traffic prediction theory to study and use spatial and temporal characteristics of short-term traffic forecasting model for short-term classification, principles established model on this basis to explore and develop appropriate evaluation model.Secondly, after the related theory of artificial neural networks to study proposed BP neural network to predict the combination of short-term traffic prediction scheme to establish short-term traffic forecast model based on BP neural network. Key to the establishment of such short-term forecasting model is that BP neural network topology to establish the parameters established in the discussion network layers model selection, determine the number of neurons, such as model predictive data processing key steps, but also for limitations of BP neural network are analyzed and discussed, optimized and improved measures proposed for its limitations.Again, for the previously proposed to optimize BP neural network weights and thresholds to avoid falling into local minimum network model this idea, the choice of adaptive genetic algorithm for global optimization search algorithm-BP establish genetic neural network model to predict short-term traffic volume. After the use of genetic algorithm to optimize the weights and the threshold network model, further training and simulation model to predict the network to improve the predictive accuracy of the model.Finally, short-term traffic data Yinchuan urban intersections measured for samples to establish two different input classification scheme based on short-term forecasting models using neural network toolbox MATLAB2009 a BP neural network traffic to establish short-term forecasting model. Genetic-BP neural network model of traffic short-term forecasting, using two different models combine programs and research Intersection Road traffic prediction, and the prediction results were analyzed and compared. The results show that, BP neural network model to predict the results basically meet the application requirements; Genetic-BP neural network model optimized to avoid the defects of BP neural network to improve the prediction accuracy, and more value in use.