基于K-means++的轨道交通异常出行乘客分类研究  

Research on Rail Transit Abnormal Travel Passenger Classification Based on K-means++Algorithm

作　　者：倪汪凌 褚文斌 NI Wang-ling;CHU Wen-bin(Shanghai University of Engineering Science,Shanghai 201620;Shanghai Shentong Metro Co.,Ltd.,Shanghai 200000,China)

机构地区：[1]上海工程技术大学,上海201620 [2]上海申通地铁股份有限公司,上海200000

出　　处：《物流工程与管理》2023年第5期96-99,共4页Logistics Engineering and Management

摘　　要：为聚焦轨道交通乘客主要出行规律,需在利用交通大数据做研究时明确异常数据和正常数据之间的界限,当研究主题、研究深度和精度要求不同时,对异常出行群体数据的接受程度也会相异。文中基于大都会二维码数据识别存在异常出行的乘客,引入K-means++聚类算法对异常出行乘客进行分类研究,并构建异常得分指标评价不同群体异常程度。聚类结果表明,将异常乘客分为5类时效果最佳。结合异常乘客出行指标、异常得分、群体占比和周出行平均频次四个角度分析群体特征,并针对性提出相应群体的数据剔除建议,以保证研究中样本的多样性和结论可靠性。In order to focus on the main travel rules of rail transit passengers,it is necessary to clarify the boundary between abnormal data and normal data when using traffic big data for research.When the research topic,research depth and precision requirements are different,the acceptance degree of abnormal travel group data will be different.In this study,passengers with abnormal travel were identified based on the QR code data of metropolitan,K-means++clustering algorithm was introduced to classify passengers with abnormal travel,and the anomaly score index was constructed to evaluate the degree of anomaly of different groups.The clustering results show that the effect is best when the abnormal passengers are divided into five categories.The characteristics of groups are analyzed from four perspectives of the travel index of abnormal passengers,the abnormal score,the proportion of abnormal passengers and the average weekly travel frequency,and the corresponding data elimination suggestions are put forward to ensure the diversity of samples in the study and the reliability of the conclusion.

关 键 词：轨道交通大数据 异常出行行为 k-means++算法 异常乘客分类 

分 类 号：F570[经济管理—产业经济]