Research On Statistical Characteristics Of Short-term Traffic Flow Data At Intersections In Mountainous City

With the progress of the times,the number of cars in China has risen sharply,and the problem of traffic congestion has become increasingly prominent.How to solve the problem has become a common problem faced by the government and the public.The most effective and reasonable way to reduce traffic congestion is traffic management and decision-making.Traffic signal control has been recognized as an economical and effective method to manage urban congestion.Traffic flow is an important basic characteristic parameter in traffic flow theory,and traffic volume analysis can ensure the sound of the traffic management system and timely monitoring of traffic conditions.So it is very important to estimate the distribution of traffic flow.At present,the existing researches mainly focus on the detection of traffic flow data or short-term traffic volume prediction,which can only represent the changes of specific standard parameters without considering the probabilistic nature,and there are not many achievements in analyzing its distribution and statistical characteristics.Therefore,this paper takes the statistical characteristics of traffic flow as the research object,and studies its probability distribution and parameter estimation.This paper first expounds the basic theory of traffic flow,and describes the data sources and characteristics of the case,from the general characteristics and statistical characteristics of two aspects of the study;This paper introduces the pretreatment method of the original traffic volume data,from how to identify to how to repair,and verifies the validity of the repair data through the case of intersection A;Then the advantages and disadvantages of several common traffic probability distribution models and different parameter estimation methods are analyzed.Then,the basic theory of maximum correlation entropy is introduced,and the maximum correlation entropy criterion is applied to parameter estimation based on the gradient rise algorithm,and the parameter estimation method based on the maximum correlation entropy criterion is constructed.The normal distribution,lognormal distribution,Weibull distribution and Rayleigh distribution are used for model fitting,and the modeling steps of the maximum correlation entropy criterion parameter estimation method are described in detail.The concrete idea of solving this method is given.Secondly,this paper selected five goodness-of-fit evaluation indexes,including the error square sum test,decision coefficient test,chi-square test,root mean square error test and comprehensive index test,to test the fitting effect,and proposed the optimal fitting effect of parameter estimation method through the case verification of intersection A.Finally,the short-term traffic volume data at intersection B was taken as a comparative case for analysis,and the applicability of the parameter estimation method based on the maximum correlation entropy criterion was studied.The parameter estimation results based on the maximum correlation entropy criterion established in this paper were compared with the maximum likelihood estimation and least square estimation results,and the evaluation index was used to test the fitting results.The case study of intersections A and B shows that the parameter estimation based on the maximum correlation entropy criterion established in this paper has the better effect and strong practicability.The parameter estimation method based on the maximum correlation entropy criterion established in this paper has a good application advantage for non-zero mean,non-Gaussian and non-stationary sequences,and improves the fitting effect of traffic flow parameter estimation to a certain extent.Through theoretical analysis and experimental verification,It is proved that the traffic flow probability distribution parameter estimation method based on the maximum correlation entropy criterion designed in this paper is accurate and feasible,verifies the statistical characteristics of traffic volume data,and can provide effective help for short-term traffic volume prediction and traffic management and control.