Testing Change Points And Its Application In Transport Hub

The change point problem is one of the important aspect in statistical research.Change point detection is used in the financial,economic,computer and other fields.In this paper,we mainly study the theory of the mean change point test of the common exponential family,the method of FDR-SMUCE,H-SMUCE and their application in urban transportation.This paper mainly studies the pedestrian data of several major subway access gates in Shanghai and the change-point inspection of passing vehicles at a certain intersection in Changchun People’s Street.It also analyzes and studies the location and time of change,and gives a reasonable solution.Use the method of segmentation to deal with the multiple change point test in practical application.And then we apply it on traffic for passenger flow,main road traffic,find the position and time of change point,and then we can infer its cause.Theoretically,this paper mainly studies the problem of the change-point test of the exponent family.Improve it on the basis of the mean test in the existing known distribution change point test and construct the mean test statistic.At the same time,give the estimate of the change point.The main research work includes the following aspects:1.We propose the method of testing change points with the exponential family distribution.Under the condition of the same variance,we test whether exist change point in the mean value,we give the test statistics under some conditions when the distribution is unknown,and prove that the asymptotic distribution is the extreme value distribution.The result is same as many literatures.This method is extended to testing multiple change points.By simulating Gamma distribution,Weibull distribution,log-normal distribution,two-parameter exponential distribution and their mixture distributions,we find the appropriate segment interval size and give the algorithm of the single change point and the multiple change points.2.We use the method of FDR-SMUCE and H-SMUCE to test the multiple change points.By simulating the multiple change points data of the ladder-like Gaussian distribution and comparing the advantages and disadvantages under two different values of σ with two different α,we choose the method of median range to determine the estimation of σ.3.We apply the mean common exponential family distribution change point test,FDR-SMUCE method and H-SMUCE method to the field of urban transportation.We select the pedestrian flow data of Shanghai Metro,People’s Square Station of Shanghai Metro and Xujiahui Station in April 1,2015,and vehicle traffic data in Renmin Road of Changchun City from January 10 to 11,2014.Multiple change points and the peak period of passenger flow is obtained.At the same time,we have judged the merits and demerits of three kinds of change point test methods.