基于PageRank的动态网络核心节点检测及演化分析  

Vital Node Detection and Evolution Analysis in Dynamic Networks Based on Page Rank

作　　者：王玙[1] 刘东苏[1] Wang Yu;Liu Dongsu(School of Economics and Management,Xidian University,Xi ＇an 710071)

机构地区：[1]西安电子科技大学经济与管理学院,西安710071

出　　处：《情报学报》2018年第7期703-711,共9页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金青年基金"大规模动态社交网络社团检测算法研究"(71401130)

摘　　要：Page Rank算法是应用最广泛的静态网络节点中心性排名算法,拓展Page Rank算法,使其能够用于计算动态网络的节点中心性是非常有意义的研究问题。本文首先基于网络重构和随机游走策略重构,定义了两种动态网络Page Rank中心性;继而给出利用分段线性拟合刻画节点中心性演化过程、预测节点中心性变化趋势的算法;最后构造图书情报领域的动态科学家合作网络,利用本文定义的中心性得到作者影响力的变化趋势,通过与真实变化趋势相比较,验证所提中心性定义的有效性。实验结果表明,本文提出的动态网络中心性能够更加准确的刻画节点中心性的演化过程、预测节点中心性的变化趋势。The Page Rank algorithm is widely used to detect vital nodes in static networks. Finding out a manner in which this algorithm can be extended to detect vital nodes in dynamic networks is a significant task. Two dynamic Page Rank centrality definitions are separately proposed based on network reconstruction and random walk policy reconstruction. Then, a segmented least squares algorithm is presented to characterize the evolution process and predict the trends of nodes＇ centrality. Dynamic co-author networks are constructed in the Library and Information fields to verify the effectiveness of our two dynamic centrality definitions. We compare the trends of author influence obtained by our methods with real-world trends. Our experiments demonstrate that using our definitions, we could characterize the evolution process and more accurately predict the trends of nodes＇ centrality.

关 键 词：动态网络 中心性 佩奇排名 随机游走 网络重构 

分 类 号：O157.5[理学—数学]