Abstract

Methods for monitoring network in the literature mainly use two approaches. One approach assumes a statistical model generated the network data and monitors the estimated parameters of this statistical model over time. The other approach summarizes a network using centrality measurements and then monitor these values. In this article, we assume zero-inflated Poisson regression as an underlying distribution to model the interactions between social actors. We show the applicability of the model in real data set. To monitor social networks based on this model we propose a Likelihood Ratio Test which examines the hypothesis of no change against hypotheses of change overtime. An EWMA chart is also developed to monitor the average degree centrality measurement. The performance of the proposed methods is investigated using simulation studies.

摘要

文献中监测网络的方法主要采用两种途径。一种方法假设统计模型生成网络数据并随时间监测该统计模型的估计参数。另一种方法使用中心性测量来总结网络，然后监控这些值。在本文中，我们假设零膨胀泊松回归作为基础分布来模拟社会参与者之间的相互作用。我们展示了模型在真实数据集中的适用性。为了基于此模型监控社交网络，我们提出了一种似然比检验，该检验检验了没有变化的假设与超时变化的假设。还开发了 EWMA 图来监控平均度中心性测量。使用模拟研究研究了所提出方法的性能。