基于旅客-航班异构网络的旅客同行子图抽取  

CO-TRAVEL SUBGRAPH EXTRACTION BASED ON PASSENGER-FLIGHT HETEROGENEOUS NETWORK

作　　者：卢敏[1,2,3] 王彦威 Lu Min;Wang Yanwei(College of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China;Civil Aviation Passenger Service Intelligent Application Technology Key Laboratory,Tianjin 300300,China;Information Technology Research Base of CAAC,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300 [2]民航旅客服务智能化应用技术重点实验室,天津300300 [3]中国民航大学信息技术科研基地,天津300300

出　　处：《计算机应用与软件》2022年第2期259-265,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61502499);天津市自然科学基金项目(18JCYBJC85100);教育部人文社会科学研究规划基金项目(19YJA630046)。

摘　　要：由于旅客-航班异构网络仅有高度稀疏的民航旅客同行记录,现有子图抽取方法难以从旅客-航班异构网络中获得旅客同行子图。对此提出基于旅客-航班异构网络的旅客同行子图抽取算法。将旅客-航班异构网络转换为旅客-旅客同构网络,通过随机游走方法得到旅客间的潜在同行关系,使用标签传播算法进行子图抽取。在国内某航空公司的旅客订票数据集上的实验表明,相比于LPA、COPRA、CPM等基准算法,该算法在模块度和标准化互信息上具有更好效果。Since there are highly sparse civil aviation co-travel records in passenger-flight heterogeneous,the existing subgraph extraction algorithms are difficult to obtain the co-travel subgraph.To solve this problem,this paper proposes a co-travel subgraph extraction algorithm based on the passenger-flight heterogeneous network.It converted passenger-flight heterogeneous network to passenger-passenger isomorphic network.And random walk was employed to discover passengers co-travel relationship.We used the label propagation algorithm for subgraph extraction.The experimental results on the passenger booking dataset records from a domestic airline show that the proposed algorithm outperformed the baseline algorithms such as LPA,COPRA and CPM in terms of modularity and normalized mutual information.

关 键 词：子图抽取 复杂网络 异构网络 旅客同行 随机游走 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]