Analysis And Application To The Statistical Characteristics Of Dynamic Road Network

With the rapid development of society economic, there are more and morecommunication and cooperation around regions. It also brings with the rapiddevelopment of traffic. But the increasingly serious traffic congestion problems makeurban residents inconvenient in daily life and daily travel. In recent years, people havecome to realize that only depend on the increasing of highway mileage may notfundamentally solve the increasing demand of travel.It has to establish an intelligenttransportation system which can make a rational use of road resources and achieve thebest flow of people and goods. To establish such a system, the basic problem is toresearch the information degeneration and complexity of the dynamic network traffic.Among them, the space and time distribution of the dynamic traffic networkinformation is one of the most important research directions of the time-varying signand the day cycle regularity of traffic information. It reflects the degeneration andcomplexity inherent regularity of the traffic information. Therefore, analyzing thespace and time distribution of the dynamic traffic network information is theprecondition of establishing an intelligent transportation system or an intelligentlogistics system.Based on this background of the subject, the paper has chosen the road trafficspeed as the attribute value of the performance of traffic characteristics. It uses hugedata of GPS dynamic floating car. And on the basis of overseas and domestic researchstatus, the paper proposes using the statistical method of Markov random field,spectral clustering analysis to analyze the two hot issues of space and timedistribution of the dynamic traffic network information. That is space and time jointdistribution of traffic information and the analysis of traffic information areadistributionBecause of slow convergence speed of EM algorithm and the algorithm is highlydependent on the initial values of the defects. After building a study model of timeand space joint distribution of the dynamic network based on Markov random field, the paper proposes a global way to solve the parameters. Firstly, it choice the PMFrecovery algorithm to fill up the missing data, then it uses the maximum likelihoodestimation to solve the model parameters in the complete data set. Further it improvesthe way to solve the parameters under the situation of the model is missing data.Finally, the paper applied each of the statistical distribution research methods in theZhuHai real network data, to know the traffic information statistical distributioncharacteristics and the regularity evolution of the section stat.It also provides theinformation and basis for grasping the network operation state overall. trafficprediction and traffic control.