Traffic State Prediction And Spatiotemporal Evolution Characteristics For Segments Of Expressway

With the increase in the number of motor vehicles and the travel demand of residents in China,the problem of expressway traffic congestion has become increasingly prominent,with frequent congestion in some segments of highways.Accurate identification and prediction of the traffic state of expressway segments and in-depth research of potential traffic state evolution rules are of great significance for active and refined management of expressways.At the same time,the traditional data acquisition methods have shown obvious limitations in the research and analysis of traffic conditions.The continual expansion of data openness provides new ideas and conditions for the acquisition of multi-source data.From the perspective of multi-source data mining,this dissertation explores the influence of different factors on expressway traffic flow,constructs the traffic state recognition and prediction model of expressway segments,and studies the evolution characteristics of traffic state between multiple segments.Firstly,the influence of different factors in the traffic environment on expressway traffic state is analyzed.Python is adopted to call the corresponding network API interface to acquire the traffic flow data and the weather data of highway segments,and the corresponding data preprocessing method is given.Based on the highway design documents,the road alignment data corresponding to each section is calculated.The data of weather,road alignment and road traffic flow are combined to analyze the influence of rainfall,visibility,cumulative curvature,and cumulative slope on the traffic flow of segments of expressways.Secondly,a combined prediction model of expressway traffic state considering traffic environmental factors is proposed.Combined with the characteristics of expressway traffic flow having strong randomness and being affected by various traffic environmental factors,the Ensemble Empirical Mode Decomposition is employed to decompose the original average travel speed sequence of expressway segments.Then the Dynamic Time Warping is applied to measure the correlation between each component and the original sequence,and the components are divided into two kinds: high correlation component and low correlation component.Then the components are combined and the traffic environment factors are taken into account to make a prediction by using short and long time memory neural networks,so as to realize the construction of the combined prediction model.In view of the difference in expressway operation in different regions,the traffic state is divided according to the normal speed and the average travel speed of sections.Multiple sections of the G65 Baomao Expressway were selected for example analysis and verification.Compared with the traditional single prediction model,the prediction accuracy of the combined model was improved in different scenarios.Considering the regional differences of expressways,the traffic state is divided according to the normal speed and the average travel speed of sections.Multiple sections of the G65 Baomao Expressway were selected for example analysis and verification.Compared with the traditional single prediction model,the prediction accuracy of the combined model was improved in different scenarios.Finally,a method for analyzing spatiotemporal evolution characteristics of traffic state between multiple segments on the basis of data mining is proposed.According to the topology of the expressway network,the spatial and temporal weight matrix of spatialtemporal Moran’s I index is improved.The spatial-temporal Moran’s I index is used to analyze the spatial-temporal correlation of traffic state between multiple segments and study the identification method of congestion-prone segments.On this basis,the FPGrowth algorithm is used to find out the evolution characteristics of traffic state between multiple segments,and the potential law of congestion propagation from the original segments to the surrounding segments is explored.Taking several segments of the G65 Baomao Expressway as an example,it is found that the segments of the G65 Banan Toll Station to Jieshi Service Area,Nanhuan Interchange to Yuxiang Interchange and other segments are the congestion-prone sections within the experimental scope according to the calculation results of the improved Moran’s I.At the same time,using the FP-Growth algorithm to discover the results of the traffic state evolution rules in the three scenarios on weekdays,at weekends,and in the holidays respectively,the propagation path of traffic congestion can be obtained.In combination with the prediction results of the traffic state of a single road section,it is possible to issue early warnings against potential congestion or congested path,changing the passive management and control of the expressways into active prevention and control.