基于最大后验估计的谣言源定位器

随着互联网的普及,信息能够通过互联网以极快的速度被传播给大众。但同时,一些虚假信息比如谣言也借助网络的级联效应泛滥成灾,因此如何在传播网络中快速准确地确定谣言传播源成为一个亟待解决的问题。文章针对社交网络提出了一种谣言源定位的方法,与现有的基于最大后验(Maximum-a-posteriori,MAP)概率估计的方法不同,该方法首先考虑全局和局部感染点、非感染点的影响,使用效果更优的MAP先验概率估计(Prior Probability Estimation,PPE)计算方式。然后基于最小生成树贪心算法来稀疏化社交网络,让MAP中的似然估计(Likelihood Estimation,LE)计算更符合真实的传播结构。最后,采用新的MAP值来估计传播网络中节点为传播源的可能性,从而更准确地定位谣言源点。将所提方法与现有的几种方法分别在模型网络和真实网络中进行了对比,实验结果表明,所提方法优于现有的谣言源定位方法。

Rumor Source Detection in Social Networks via Maximum-a-Posteriori Estimation

With the popularization of Internet,information can be transmitted to the public at an extremely rapid rate through Internet.But at the same time,some abnormal information,such as rumors,has been flooded with the cascade effect of Internet.How to quickly and accurately identify the source of a rumor spreading under a complex network becomes an urgent problem to be solved.This paper proposes a source localization algorithm in social networks.Different from some existing methods based on Maximum-a-posteriori(MAP)probability estimation,this method first considers the influence of global and local infected nodes and non-infected nodes,and proposes a better MAP prior probability estimation(PPE)calculation mode.Then,asocial network is sparsified through a greedy algorithm based on minimum spanning trees,which makes the likelihood estimation(LE)calculation in MAP more consistent with the real propagation structure.Finally,a new MAP value is used to estimate the possibility of a node as the source of propagation in the social network as to locate the source of the rumor more accurately.The proposed me-thod is compared with some existing methods by an experiment on some model networks and real networks,and experimental results show that the proposed method is superior to these existing methods of locating the rumor source.