Data Quality Controlling And Evaluating Method For Traffic Flow On Highway

With the rapid growth of highway travel demand,the traffic congestion problem and traffic safety problem continue to arise.In order to improve highway traffic conditions and safety performance,the Intelligent Transportation System(ITS)emerged.The core of the Intelligent Transportation System is to support the traffic management decision and travel information service using high-quality traffic flow data collected,and the quality of traffic data directly affects the formulation and implementation of the traffic management decision.Therefore,it is necessary to study the evaluation system and control methods on highway traffic flow data quality.In this paper,focusing on the problem of highway traffic flow data,a traffic flow data quality evaluation and control method is established to improve data quality and provide fully support for highway traffic management and transportation planning.Firstly,the domestic and overseas research process in data quality evaluation and data quality control is summarized,with the data quality evaluation and control methods set as the study objects of the paper,then the research content and technical route of the paper are presented.Basing on the research objects,the space-time distribution characteristic is analyzed using the actual highway traffic flow,providing a theoretical support for the evaluation and control of traffic flow data quality.Secondly,according to the theory of three parameters of traffic flow and the analysis of data characteristics,the types of the traffic flow data problems are determined.For different types of data quality problems,an identification method on traffic flow data quality problems is presented.From the aspects of data integrity and validity,the identification method follows a logical order of nonnative detection,missing data detection,error data and inaccurate data detection.Using the collected traffic flow data of Dujiakan section of Beijing-Hong Kong-Macao Highway,the proposed identification method is verified.Thirdly,based on the analysis of highway traffic flow data quality,a traffic flow data quality evaluation index system that combines dot,line and surface is proposed from the microscale,mesoscale and macroscale levels.The evaluation index system is used to describe the data quality of highway traffic flow in three different dimensions:road section,route and road network.Combined with evaluation system,the analytic hierarchy process(AHP)-fuzzy comprehensive evaluation method is established to evaluate the quality of highway traffic flow data in different dimensions of points,lines and surfaces.From the different dimensions of points,lines and planes,he highway G4 Dujiakan section,highway G4 route,Shandong Province highway network are selected as the research objects correspondingly,and an data quality empirical study is conducted in microscale,mesoscale and macroscale levels with the data quality evaluation methods proposed.The results show that the proposed evaluation index system and method can realize the three-dimensional highway traffic flow data quality evaluation.Finally,based on the identification and evaluation of the highway traffic flow data quality,a gradient descent Tucker decomposition data quality control method is proposed with the introduction of the multidimensional tensor analysis decomposition.The data quality method integrates the multi-dimensional temporal and spatial characteristics of the highway traffic flow data to improve the traffic flow data quality.Using the actual speed data of highway G4 Dujiakan section,three tensor formations are constructed,and the data quality is controlled.The second evaluation of data quality in different dimensions of points,lines and surfaces is realized with the proposed data quality control method.The research results show that the proposed data quality control method can repair traffic flow data quality problems and improve data quality evaluation results,while different tensor forms of the same data construction have different data control effects.