Research On Traffic Model Based On Density Clustering And Bayesian Estimation

Traffic discrimination and prediction has been the important one of intelligenttransportation links, and the discriminant condition must rely on traffic parameters, highwaytraffic main parameters are traffic flow,density,velocity,etc;These parameters are from thetraffic data to be detected, it must rely on the detector. At present, there are many mature datacollection methods, but also the advantages and disadvantages exist; In view of the traditionaltesting equipment installation cost is high, maintenance costs, and easily damage,affected bythe natural conditions of climate, the application of the mobile phone positioning technologyacquiring traffic data, due to the characteristics of small investment,covering range, receivedattention at home and abroad; the location of the phone now is difficult to meet the precisionof the construction of ITS requirements,but it still has an important development trend.Thispaper assume that the accuracy meet the condition. Firstly,obtain the cell phone positioningdata, and process data to form vehicle samples, and then use the density cluster to divide intoseveral regions,extract the regional traffic parameter then discuss,finally get the trafficparameters for fuzzy reasoning,at the same time use Bayesian estimation constantly to modifythe membership function in the fuzzy reasoning and thus to identify the traffic condition;This paper make use of the mathematical statistics and data mining to provide a new approachand idea for the traditional technology and methods of traffic discrimination.on the basis ofthe mobile positioning in accordance with the data analysis—the theoretical derivation—data-examples of ways to related research.mainly work:1) introduce the mobile phone positioning in data acquisition, considering thecharacteristics of vehicle-mounted mobile phone,change mobile phone sample into vehiclesample,data processing for traffic parameters(vehicle sample and speed)2) when a certain period is not smooth, showing the shape of traffic is often not spherical,sowe use the density cluster in vehicle sample and then process the density clustering resultsanalysis and discrimination (analysis every type of vehicle number is more than threshold ornot to process discrimination) and turn the extraction of threshold to discussion; In addition,use example to verify the density cluster and k-means clustering comparison;3) the threshold is seen as the unknown parameter, estimates of the adjusted at this time, so the bayesian estimation, aims to use the effect of prior distribution and sample distribution,sothe threshold is more reasonable. Assume that the traffic flow separately to random poissondistribution and binomial distribution when smooth and Congestion. Firstly use thechi-square goodness-of-fit of hypothesis test base the two distribution, use bayesianestimation on the threshold, and also compare with the maximum likelihood estimation.4) according to the threshold obtained, determine the fuzzy division of trafficcorresponding membership functions, fuzzy process the sample data, and establish the fuzzydecision tree, extract the fuzzy rules, with the extraction of the rules for the test sample offuzzy reasoning, to realize the traffic discrimination and prediction. The experiment result hasgood accuracy...