Prediction For Short-term Traffic Flow Based On Optimized Wavelet Neural Network

Intelligent traffic management system is the important means to solve the problem of modern transportation, due to complex changes of the urban road traffic system, The prediction is very difficult. As the basis of intelligent transportation system, short-term traffic flow prediction is of great importance, therefore, the key research of this article is about traffic flow prediction model improvement based on analysis of the relevant basic knowledge, which requires historical and real-time traffic flow data collection and processing, purpose is to provide basic data and decision support intelligent traffic management system research, making traffic management decision-making more scientific.Traffic flow is affected by complex external factors, has high characteristics with complexity and nonlinear. There is no single prediction model can satisfy the high requirement, so in this paper, the predictive model was improved. and the main work has three aspects as follow:Firstly, basic definition and nature of the traffic flow was analyzed, and traffic flow data collection methods was expounded, the abnormal data identification method was analyzed, and also the four types of traffic flow prediction model was introduced.Secondly, this paper studied neural network, a wide application of traffic flow prediction model, based on this, the improved particle swarm algorithm was put forward to optimize the wavelet neural network model for forecasting the traffic flow. In the light of the wavelet neural network is extremely strong nonlinear processing ability, self-organizing, adaptive and learning ability, and use it as a basic model, however, wavelet neural network also has shortcomings such as slow convergence speed, so taking the advantage of particle swarm algorithm, such as convergence speed, high robustness, and strong global search ability, particle swarm optimization algorithm was used to train the parameters of wavelet neural network training, so particle swarm algorithm of wavelet network forecasting model was formed, to improve performance of the basic wavelet neural network model in the short-term traffic flow prediction. Finally, further research was down to improve particle swarm optimization(pso) algorithm, the improvement focus on key parameters of particle swarm optimization(pso) algorithm. Since cloud model has great advantage on the swarm intelligence algorithm, particle swarm optimization(pso) algorithm was optimized using cloud model, and then searching and optimization ability is improved, thus wavelet neural network model of short-term traffic flow was established based on the cloud theory.At the end, the basic swarm intelligence algorithm and the improved swarm intelligence algorithm was compared, also did the advantages and disadvantages of wavelet neural network model and the optimized one, when input data in the trained model produce prediction Error.