Research On Traffic Status Identification And Prediction Method For Urban Expressway

With the urban expressway traffic volume continues to grow,more and more problems such as traffic congestion,traffic accidents.It is far from enough to rely only on the increase of roads and other infrastructure to ease the pressure of traffic travel,intelligent traffic management and information-based traffic travel guidance for urban expressways should be considered.With the acceleration of ITS construction in large and medium-sized cities,a large number of traffic detectors installed on urban expressways provide sufficient basic traffic flow parameter data,which provides an important data source for the study of urban expressways traffic status identification and prediction methods.Through the analysis of traffic flow parameter data,mining effective traffic characteristic information,reasonable traffic condition assessment and prediction method is put forward,which provide more for traffic managers and traveler timely,reliable and predictable traffic management and transportation decision-making,has become the effective measures to alleviate the traffic problems for urban expressway.This paper carries out research on the identification and prediction methods for urban expressway traffic status,and the specific research work and research results mainly include the following aspects.The characteristics for urban expressway traffic flow parameter data are summarized from two aspects of main traffic flow parameter data and auxiliary traffic flow parameter data,and the reasons and types of missing traffic flow parameter data are clarified.Based on the Coupled Matrix and Tensor Factorizations(CMTF)model,the missing data repair method for urban expressway traffic flow was constructed.By constructing weather feature matrix and lane feature matrix from auxiliary traffic flow parameter data and coupling them to tensor model established from main traffic flow parameter data,the Alternating Direction Method of Multipliers(ADMM)algorithm is used to optimize the objective function.The result of empirical analysis shows that CMTF method has better repair effect.Based on the analysis of the variation characteristics of the traffic flow parameters before and after the traffic incidents,the incident variables were constructed from multiple perspectives,and the characteristic variables that were more sensitive to the traffic incidents were extracted by the Factor Analysis(FA)method,which were used as the final input incident variables of the AID algorithm for the urban expressway.Based on the Random Forest algorithm(RF),the Weighted Random Forest(WRF)AID method was constructed for urban expressway.The Bootstrap re-sampling algorithm with improved constraints was used to effectively extract the initial training data set and complete the training.Matthews Correlation Coefficient(MCC)coefficient is used to evaluate the classification effect of decision trees in random forest after training and is assigned to each decision tree as a weight,so that the tree with better classification ability has more voting rights in the voting stage.The empirical analysis shows that FA-WRF model can obtain effective detection results.Combined with the existing classification standard,the traffic status of urban expressway of urban expressway is divided.Based on the availability and usability,the status variables are set based on three basic traffic flow parameters:flow rate,speed and occupancy rate.On the basis of Fuzzy C-Means clustering(FCM)algorithm,the traffic status identification method of urban expressway was designed based on DSFA-FCM.The initial clustering center of FCM algorithm was determined based on Dynamic Step Firefly Algorithm(DSFA),and then the traffic status identification of urban expressway was realized by using FCM algorithm.The empirical results show that the DSFA-FCM model can effectively recognize traffic status.In view of the traffic flow parameter data of urban expressway with strong nonlinear characteristics,the phase space reconstruction of multi-variable traffic flow parameters is carried out by C-C algorithm,and the chaos of traffic flow parameters is judged by using Small Number method,so as to determine the input variables of short-time traffic flow parameter prediction model.In the aspect of prediction model construction,based on the Elman neural network,the MPSR-AR-Elman neural network based on the dynamic Elman neural network model was built to predict the short-term traffic flow parameters for urban expressway,the number of hidden layer nodes in the Elman neural network is effectively determined by the Advance and Retreat(AR)method.The empirical results show that the prediction effect of the MPSR-AR-Elman neural network model is effective.Involved in this paper,research contents,research methods and get the final conclusion is that the existing status of urban expressway traffic identification and prediction method of complement,perfect and further to explore,from work to further improve its urban expressway traffic safety and improve the traffic operation efficiency has important academic significance and application value.