基于组合模型的车辆出行特征模式划分  

Classification of Vehicle Travel Feature Modes Based on Combined Model

作　　者：蔡晓禹[1,2] 吕亮 杜蕊 CAI Xiao-yu;Lü Liang;DU Rui(School of Transportation,Chongqing Jiaotong University,Chongqing 400074,China;Chongqing Key Laboratory of Traffic System&Safety in Mountainous Cities,Chongqing 400074,China)

机构地区：[1]重庆交通大学交通运输学院,重庆400074 [2]山地城市交通系统与安全重庆市重点实验室,重庆400074

出　　处：《公路交通科技》2021年第6期129-140,共12页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(61703064);重庆市高校优秀人才支持计划项目;重庆市技术创新与应用示范专项重点研发项目(cstc2018jscx-mszdX0085)。

摘　　要：精准掌握车辆的出行规律研究智能化城市交通管理及规划的基础工作,而掌握车辆出行规律的前提是探究车辆的出行特征。为研究城市道路交通车辆的出行特征模式,通过对历史RFID轨迹数据挖掘,对私家车、出租车样本轨迹数据进行定性分析,总结车辆运行的分布特征规律。基于数理统计分析,建立了出行频次、在网时间、轨迹重复率、出行时段,活动偏好区域、干线影响区偏好等出行特征指标体系。通过对出行特征指标的定制选取,建立基于密度峰值(CFSFDP)算法与BP神经网络算法的出行特征群体辨识模型。研究了私家车、出租车存在的特征群体,辨识出不同的出行模式,即实现出行特征群体的辨识。选取重庆市主城区域内的RFID数据进行试验分析,分别基于私家车、出租车提取的出行特征指标,进行CFSFDP算法的聚类分析,找到聚类中心,归纳分类数据。再利用分类数据进行BP神经网络训练学习,评价模型试验结果。结果表明:私家车存在3种出行特征群体:商用私家车群体、通勤私家车群体、其他私家车群体,群体识别率为97.2%。出租车具有2种出行特征群体:其他区域偏好出租车群体、干线影响区偏好出租车群体;群体识别率高达99.18%。Accurately mastering the travel rule of vehicles is the basic work of intelligent urban traffic management and planning,and the premise of mastering the travel rule of vehicles is to explore the characteristics of vehicle travel.In order to study the travel characteristics of urban road traffic vehicles,by mining historical RFID trajectory data,the trajectory data of samples of private cars and taxis are qualitatively analyzed,and the rule of vehicle operation distribution characteristics is summarized.Based on mathematical statistical analysis,the travel characteristic indicator system which includes travel frequency,online time,trajectory repetition rate,travel period,activity preference area and mainline influence area preference is established.Through the customized selection of travel characteristic indicators,the travel characteristic group recognition model based on the peak density(CFSFDP)algorithm and the BP neural network algorithm is established.The characteristic groups that exist in private cars and taxis are studied,and different travel modes(different travel characteristic groups)are recognized.Selecting the RFID data of the main urban area in Chongqing for experimental analysis,the cluster analysis is performed based on the travel characteristic indicators extracted by private cars and taxis by the CFSFDP algorithm to find out the cluster center and summarize the classification data.Then,the BP neural network training and learning is conducted by using the classified data,and the result of model experiment is evaluated.The result shows that(1)there are 3 travel characteristic groups for private cars:commercial private car groups,commuter private car groups,and other private car groups,and the group recognition rate is 97.2%;(2)there are 2 travel characteristic groups for taxis:taxi groups with preferences in other areas,taxi groups with preferences in mainline influence areas,and the group recognition rate is 99.18%.

关 键 词：城市交通 RFID数据 出行特征指标 群体辨识 CFSFDP&BP组合模型 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]