基于集聚系数的链路预测算法

Link Prediction Based on Clustering Coefficient

复杂网络中的链路预测是指基于已知的网络结构信息来预测网络中尚未链接的两个节点间产生连边的可能性。现有算法主要是基于局部信息的相似性算法,即针对共同邻居的数量、共同邻居的度值以及共同邻居之间的相互链接程度进行研究,应用范围有限。为此,针对一个节点的邻接点之间相互链接的程度,本文提出一种基于集聚系数的新算法。本文利用该算法对多种现实网络以及pajek生成的模拟网络进行了实验,实验结果表明,该算法适用范围广,链路预测准确率高。

Link prediction in complex network aims at estimating the likelihood of the existence of links be-tween nodes by the known network structure. Currently, most link prediction algorithms are si-milarity algorithms based on local information including the number of common neighbor nodes, degree of common neighbor nodes and the interactions between common neighbor nodes, and thus their applied range is limited. In this paper, we consider the interactions between adjacent nodes of a node and design a new algorithm based on clustering coefficient. We use this new algorithm in the experiments on real networks and simulative networks generated by pajek, and experimental results show that the algorithm is applicable to a wide range of problems and it has the high accuracy of prediction.