Research On Urban Road Traffic Flow Prediction Based On Deep Belief Network

With the rapid development of society and economy,the rapid increase of automobile ownership has led to the increasing contradiction between supply and demand of urban traffic system,so that the construction measures of urban roads are far from satisfying their increasing demand for travel,and then there were all kinds of traffic problems.The application of intelligent traffic system plays a key role in the dynamic traffic management of urban roads,and traffic flow prediction,as the core technology in ITS,effectively alleviates the serious traffic problem.Therefore,the scientific and accurate prediction of urban road traffic has academic value and practical significance.In view of the complexity and high uncertainty of traffic flow time series data analysis in this research field,apply the advantage of deep learning in the analysis of non-structured time series data and propose an improved method of urban road traffic flow prediction based on deep belief network in this paper.The method also addresses the slow convergence rate of restricted Boltzmann machine input only,which leads to data loss and slow convergence in training of network model parameters,an urban road traffic flow prediction method based on deep belief network with adaptive learning step is proposed.Firstly,a Data pre-processing method is proposed which converts traffic flow time series data into non-structured time series data.Then,the improved deep belief network for pre-treated non-structured data is used to train the prediction method and traffic flow.The predictive model of this paper is to combine restricted Boltzmann machine and continuous restricted Boltzmann machine for model unsupervised learning,and BP neural network is used to monitor the learning of the model.After testing and adjusting the model parameters continuously,the prediction error of the whole prediction model can be converged to train the prediction model,and the traffic flow can be predicted by using the pre-processed non-structured time series data.Finally,the traditional deep belief network and the improved deep belief network are used for non-structured traffic flow data prediction,the experimental results show that the urban road traffic flow prediction model,which is the improved deep belief network with adaptive learning step,has higher prediction accuracy and can effectively handle the problem of urban road traffic flow prediction.