一种基于交通态势的扇区聚类分析方法  

A Clustering Analysis Method for Sectors Based on Traffic Situation

作　　者：丛玮 郑洪峰 朱睿 CONG Wei;ZHENG Hong feng;ZHU Rui(Vari Flight Technology Company Limited,Hefei 230000,China;Big Blue Hole(Nanjing) Technology Company Limited,Nanjing 210001,China;College of Civil Aviation,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)

机构地区：[1]飞友科技有限公司,合肥230000 [2]大蓝洞(南京)科技有限公司,南京210001 [3]南京航空航天大学民航学院,南京211106

出　　处：《航空计算技术》2019年第4期17-21,共5页Aeronautical Computing Technique

基　　金：国家自然科学基金面上项目资助(61773203)

摘　　要：为了提升对交通态势的认知理解,加强空域精细化管理能力,研究了基于交通态势的扇区聚类分析方法。借鉴已有研究成果,建立了航班分布和航班动态两类指标定量描述交通态势。利用主成分分析方法提炼指标信息;采用K-medoids聚类算法对多个扇区进行聚类分析;利用Dunn/DB指标评价聚类质量,确定最佳聚类数量。实例表明,所选指标可以较好地反映交通态势,基于主成分分析法可以提炼原始指标90%以上的信息,聚类分析方法有效识别了15个扇区的交通态势特征。In order to improve the cognition and understanding of traffic situation and enhance the capability of meticulous management in airspace,a clustering analysis method based on traffic situation is studied.Two categories of metrics are constructed to describe traffic situation quantitatively according to present research.Two categories of metrics focus on flight distribution and flight dynamics respectively.Primary component analysis is used to refine metrics information.Then multiple sectors are divided into different clusters by K-medoids clustering algorithm.The Dunn/DB indicator is established to evaluate clustering results and determine the optimal clustering number.The case study shows that the selected metrics can well reflect the traffic situation.More than 90% of originalmetrics′ information is extracted through primary component analysis method.The clustering method can identify the traffic situation pattern of 15 sectors effectively.

关 键 词：航空运输 交通态势 聚类分析 扇区 效能评估 

分 类 号：V355[航空宇航科学与技术—人机与环境工程]