基于向量机多分类模型的城市交通状态演化特性  

Urban traffic state evolution characteristics based on vector machine multi-classification model

作　　者：张开文 冯阳 王岚 朱云波 王龙健 ZHANG Kaiwen;FENG Yang;WANG Lan;ZHU Yunbo;WANG Longjian(Yunnan Transportation Research Institute Co.,Ltd.,Kunming,Yunnan 650011,China;School of Civil and Transportation Engineering,Ningbo University of Technology,Ningbo,Zhejiang 315211,China;School of Transportation Engineering,Chang’an University,Xi’an,Shaanxi 710064,China)

机构地区：[1]云南省交通科学研究院有限公司,云南昆明650011 [2]宁波工程学院建筑与交通工程学院,浙江宁波315211 [3]长安大学运输工程学院,陕西西安710064

出　　处：《公路交通科技》2025年第4期8-16,共9页Journal of Highway and Transportation Research and Development

基　　金：云南交投集团科技项目(YNJTKFB2023-000499);云南省交通科学研究院有限公司科研项目(JKYZLX-2023-26)。

摘　　要：【目标】为实时准确识别城市路网交通流状态,以期为缓解城市区域交通拥堵问题提供理论依据。【方法】采用支持向量机多分类算法(SVM)识别交通流状态,基于马尔可夫链构建交通流状态迁移概率模型,以西安市雁塔区的部分区域路网出租车GPS数据为基础,通过工作日和节假日对比分析,深入剖析交通流运行状态的定性定量特征及其时空变化规律。【结果】标定后的主干路SVM模型识别精确度为96.875%,次干路SVM模型识别精确度为93.519%,满足识别精度要求。工作日雁塔区部分区域主干路和次干路的拥挤流状态均更易向外迁移,其内部迁移概率分别为51.19%和43.95%,小于其他交通流状态的内部迁移概率。而节假日主干路稳定流状态更易向外迁移,其内部迁移概率为50.87%,小于其他交通流状态的内部迁移概率。工作日主干路主要处于稳定流状态,占比30.89%,次干路主要处于畅通流状态,占比54.49%。节假日主干路和次干路主要处于畅通流状态,占比分别为30.43%和54.68%。【结论】在没有外界干扰的前提下,研究区域的交通流状态较为稳定,呈现逐步迁移,较少发生跳跃迁移。当发生突发事件后研究区域的交通流状态发生明显的变化,会产生跳跃迁移。该研究可用于判别常发性和偶发性交通拥堵,为城市交通的控制、诱导、管理和决策提供有力支持。[Objective]Accurately identify the real-time traffic flow state of urban road networks,and provide the theoretical basis for alleviating traffic congestion in urban areas.[Method]The support vector machine(SVM)multi-classification algorithm was used to identify traffic flow state.A probability model of traffic flow state migration was constructed based on Markov chain.A comparative analysis on weekdays and holidays was carried out based on the taxi GPS data of some regional road networks in Yanta District,Xi’an,China.The qualitative and quantitative characteristics of traffic flow state and its spatio-temporal variation patterns were analyzed in depth.[Result]The calibrated SVM model recognition accuracy is 96.875%for arterial roads,and 93.519%for sub-arterial roads,which satisfies the recognition accuracy requirement.The congested flow states are both more likely to migrate outward on both arterial and sub-arterial roads in some areas of Yanta District on weekdays.Their internal migration probabilities are 51.19%and 43.95%respectively,which are smaller than the internal migration probabilities of other traffic flow states.While the steady flow state of arterial roads on holidays is more likely to migrate outward.Its internal migration probability is 50.87%,which is smaller than the internal migration probability of other traffic flow states.On weekdays,the arterial roads are mainly in a stable flow state,accounting for 30.89%;and the sub-arterial roads are mainly in a free flow state,accounting for 54.49%.On holidays,the arterial roads and sub-arterial roads are mainly in the free flow state,accounting for 30.43%and 54.68%respectively.[Conclusion]Under the premise of no external interference,the traffic flow state in the study area is more stable,showing gradual migration,and less jump migration.When an emergency occurs,the traffic flow state in the study area changes significantly,and jump migration occurs.The study result can be used to distinguish recurrent and incident traffic congestions,providing effecti

关 键 词：智能交通 交通流状态 支持向量机 城市交通状态 出租车GPS 

分 类 号：U121[交通运输工程]