Study On Traffic State Evaluation And Association Rules Mining On Arterial Roads

The assessment of urban road traffic status can provide the traveler with information about road network operation status and help the traveler to choose the travel route.The association rules mining for traffic conditions and finding potential laws provide a theoretical basis for traffic decision-making.At present,there is a low accuracy of the estimation of traffic parameters in the related research.The assessment model of traffic operation status has low rationality and weak sensitivity to congestion.The state dimension of association rule mining model is mostly univariate and cannot be fully capture traffic state problems.Based on the summarization and analysis of existing traffic parameter estimation methods,traffic operation state models,and association rules mining models,this paper contributes as follows:Considering the accuracy of the traffic parameters and the low-frequency characteristics of the trajectory data of floating car,the method of estimating the average speed of the interval based on the weight of the trajectory covering the road section is proposed.On the basis of estimating the average speed of the interval,the average delay of the road section and the average delay at the intersection are estimated.Using the maximum queue length as an indicator,the intersections are dynamically divided.Using the fuzzy comprehensive evaluation theory,the maximum queuing length,the average speed of the section,the delay of the average of the road section,and the average delay of the intersection are selected as evaluation indicators from the perspective of easy availability and accuracy of the evaluation index and the sensitivity of the model to congestion.Considering the complexity and fuzziness of traffic flow,the golden section method is used to divide the fuzzy range of evaluation indicators,establish the membership function,a fuzzy relation matrix,and finally a evaluation model.The traffic operation status evaluation result is the status dimension,and the spatial dimension and time dimension are added to the association rule mining.The Apriori algorithm is improved by using the classic association rules.State filtering conditions and space-time constraint conditions are added before the connection step,which reduces the space-time complexity and improves the algorithm mining efficiency.This article compares the estimated and ground-truth value of max queue length of Nanhu Road(Yan'an Avenue to Renmin Street)on June 20,2018 in Changchun City.The average relative error of the max queue length is 12.48%.The absolute error is 43 m.In order to verify the proposed mining model,traffic condition of arterial streets in Changchun City on five consecutive Friday(training sets)and Saturday(testing sets)were used for congestion spatio-temporal association rule mining from March,2017.The neighboring per iod,the shortest path distance is 5km are adopted as the spatial and temporal constraints respectively and both support and confidence are 100%.With the training set and test set,the same propagation rules were mined.The model has good applicability.