基于有向频繁子图挖掘的移动性模式网络构建方法  

Method for constructing mobility pattern network by mining directed frequent sub⁃graph

作　　者：张海涛[1] 李济平 罗城 冀康 沈慧娴 ZHANG Haitao;LI Jiping;LUO Cheng;JI Kang;SHEN Huixian(College of Geographic and Biologic Information,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;College of Telecommunications&Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学地理与生物信息学院,江苏南京210023 [2]南京邮电大学通信与信息工程学院,江苏南京210003

出　　处：《南京邮电大学学报（自然科学版）》2021年第2期62-70,共9页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(41201465);江苏省自然科学基金(BK2012439);江苏省社会发展项目(BE2016774)资助项目。

摘　　要：挖掘大量移动轨迹数据获取移动性知识,可为城市交通、智慧医疗等众多行业领域提供辅助决策。但现有的移动性知识表达方式过于简单,不能反映产生移动轨迹数据复杂系统的潜在运行规律,需要从网络的视角分析移动性知识之间的复杂关系。提出了一种基于有向频繁子图挖掘的移动性模式网络构建方法,包括移动轨迹数据到轨迹有向图的转换,基于有向图的移动轨迹频繁子图挖掘,以及基于GraphX图处理框架的移动性模式网络构建。实验结果发现:基于提出方法构建的移动性模式网络,可以使用源节点、目标节点的数量,平均节点度值(包括平均节点入度、平均节点出度)以及网络聚集系数等网络特征,清晰表达移动性知识之间的复杂关联关系。Mining mobility patterns from large scale of trajectories can provide auxiliary decision⁃making for many relevant industry applications,such as urban transportation and digital wisdom medical system,etc.However,the existed definitions of mobility patterns are too simple to reflect macroscopic characteristics of the complex geographic systems generating the patterns.Furthermore,it is necessary to analyze the complex relationships of human mobility patterns from a network view.A method for constructing a mobility pattern network by mining directed frequent sub⁃graphs is proposed,including converting the trajectories to a set of directed graphs,mining frequent sub⁃graphs from the directed graphs and constructing a set of mobility pattern networks by connecting the frequent sub⁃graphs with common nodes.The method is implemented in the big data computing platform:Apache Spark GraphX.Experimental results show that the method can effectively construct mobility pattern networks from large scale of trajectories,and the complex relationships of mobility patterns can be clearly expressed by the network characteristics,such as the number of source nodes and target nodes,the average node degree value and network aggregation coefficient.

关 键 词：移动轨迹 轨迹有向图 频繁子图 移动性模式网络 

分 类 号：TP3-05[自动化与计算机技术—计算机科学与技术]