Research On Short-term Traffic Flow Prediction And Guidance Method Based On Urban Road Traffic Data

Intelligent Transportation System(Intelligent Transportation System,ITS)refers to the establishment of a real-time,accurate and efficient comprehensive transportation management system that effectively integrate and apply advanced information technology,data communication transmission technology,electronic sensing technology and computer technology.With the rapid transmission of information in today's society,the number of data grasped and processed in intelligent transportation systems has increased exponentially,but how to analyze and master the reliable traffic information from the high-speed transmitted digital information is of far-reaching significance to traffic manager.Therefore,it is necessary to study and explore new methods for traffic state prediction,in order to make better use the rich traffic information,and further improve the accuracy and reliability of traffic state discrimination and prediction.Combining different information processing methods in modern information technology,the thesis discusses some reliable analysis methods of traffic data from the perspectives of data pretreatment,traffic prediction and traffic guidance to solve the problems of missing individual samples in traffic data,low accuracy of traffic prediction and low utilization efficiency of guidance information.The specific work content is as follows.(1)Traffic data completion method based on low rank matrixIn order to analyze and master reliable traffic information from traffic big data,combined with modern information technology,this paper discusses the characteristics of urban road traffic data and the causes of fault data.Because the fault data will seriously affect the performance of traffic information system,this paper studies the completion of traffic data.Firstly,this paper introduces the low rank matrix completion model based on kernel norm minimization;secondly,in the traditional low rank matrix completion method,the singular value part and the minimum norm are replaced by the kernel norm to recover the traffic data of the low rank matrix;finally,an improved method of introducing the ordered constraint into the singular value part and minimizing is proposed.Some missing traffic data is recovered by proposed method and verified by two real traffic data,which proves that the proposed method is superior to the traditional method(2)Short term urban road traffic flow prediction method based on multi-traffic parameters fusionAiming at the chaos phenomenon of traffic system,in order to reflect the changing characteristics of traffic conditions comprehensively and better,multi traffic parameters are used to provide complete traffic information for short-term traffic prediction from different aspects.First of all,because the one-dimensional time series structure is single and contains little information,which cannot show the motion rule of high-dimensional complex system.Therefore,according to the chaotic characteristics of traffic parameters,the short-term traffic flow prediction model needs to introduce phase space reconstruction and forecast according to the objective rule after phase space reconstruction;secondly,the phase point fusion of multi parameter time series in high-dimensional phase space based on bayesian estimation theory is studied;thirdly,the traffic state of multi parameter based on maximum Lyapunov exponent is introduced.Finally,the general regression neural network is applied to improve the multi-parameter traffic state prediction model based on the maximum Lyapunov exponent,and the accuracy of the improved prediction method and the traditional prediction method based on Lyapunov exponent is compared and verified by using the urban road traffic data.The results show that the improved multi-parameter traffic state prediction model based on the radial basis function neural network is effective and the traffic state prediction method has high prediction accuracy.(3)Short term prediction method of urban road network traffic flow based on multi-neural network fusionThe short-term traffic flow prediction method based on multi-traffic parameters fusion of general regression neural network is easy to fall into the local minimum state in the case of complex road network.Through the neural network algorithm to identify the characteristics of complex nonlinear system and the mutual complement of various neural network models,a fusion of radial basis function neural network,rerrent neural network and generalized regression neural network is proposed.Through the complementary advantages and disadvantages of various models of neural network,it is verified that the prediction results after fusion can better improve the short-term prediction accuracy of urban road traffic data.(4)A new method of VMS(Variable Message Sign,VMS)optimal deployment based on maximizing the utility of induced informationOn the basis of the existing model of maximizing induced utility,a new method for optimal layout of variable information boards based on the maximization of actual induced utility is proposed.This method is mainly to improve the original utility maximization model.By analyzing the complexity of the induction utility of the variable information board,the utility of repeated induction and wasted utility is added to the calculation of actual utility,and the induction coverage redefine the induced repetition rate and finally design a function solution method based on greedy algorithm for optimal layout of the information board.Through the network examples of 36 road sections,the method is verified to be simple and effective.The information board can be optimized by analyzing the complex conditions of traffic flow,and the guidance of the entire system can be improved under the conditions that the regional road complex conditions and traffic demand points are relatively determined.Therefore,the proposed VMS layout is more in line with the needs of actual traffic flow guidance.(5)VMS guidance information release method based on bi-level programming modeIn order to reasonably induce urban traffic flow.A method of guidance information release based on VMS is proposed to improve the overall operation efficiency of the road network.Firstly,the process of game between traffic system managers and users is analyzed.Secondly,the optimization model of VMS guidance information release strategy is established based on the dual-programming model.Finally,the genetic algorithm is used to solve the optimization model.The effectiveness of the optimization process of VMS guidance strategy is verified through the real road network.