Tracking changes in dynamic information networks
Tracking changes in dynamic information networks
复制标题
DOI：
10.1109/cason.2011.6085925
复制
发表时间：
2011-12
期刊：
影响因子：
--
通讯作者：
M. Takaffoli;Justin Fagnan;Farzad Sangi;Osmar R Zaiane
中科院分区：
文献类型：
--
作者：
M. Takaffoli;Justin Fagnan;Farzad Sangi;Osmar R Zaiane
社会网络分析是一门新兴的学科，用于分析社会结构和信息网络，以揭示网络中顶点之间的交互模式。大多数社交网络都是动态的，研究这些网络随时间的演变可以深入了解图中节点所表达的个人行为以及它们之间的信息流。在动态网络中，社区是一组密集互连的节点，会受到底层人口变化的影响。对社区及其演化的分析可以帮助确定网络结构特性的变化。我们提出了一个框架，随着时间的推移建模和检测社区的演变。首先，我们提出的社区匹配算法有效地识别和跟踪类似的社区随着时间的推移。然后，一系列的重大事件和过渡的定义来表征网络的演化，其社区和个人。我们还提出了两个指标，称为稳定性和影响力指标来描述个人的积极行为。我们目前的实验，探索社区的动态安然电子邮件和DBLP数据集。使用从检测到的社区中提取的主题来评估事件表明，我们可以在真实的数据集中随着时间的推移成功跟踪社区。
Social network analysis is a discipline that has emerged to analyze social structures and information networks to uncover patterns of interaction among the vertices in the network. Most social networks are dynamic, and studying the evolution of these networks over time could provide insight into the behavior of individuals expressed by the nodes in the graph and the flow of information among them. In a dynamic network, communities, which are groups of densely interconnected nodes, are affected by changes in the underlying population. The analysis of communities and their evolutions can help determine the shifting structural properties of the networks. We present a framework for modeling and detecting community evolution over time. First, our proposed community matching algorithm efficiently identifies and tracks similar communities over time. Then, a series of significant events and transitions are defined to characterize the evolution of networks in terms of its communities and individuals. We also propose two metrics called stability and influence metrics to describe the active behavior of the individuals. We present experiments to explore the dynamics of communities on the Enron email and DBLP datasets. Evaluating the events using topics extracted from the detected communities demonstrates that we can successfully track communities over time in real datasets.