基于非负矩阵分解的复杂网络重构

摘要

将网络连边的产生机制和其社团结构结合在一起，基于社团结构决定网络连边的假设推导出节点间的连接概率矩阵并表达为矩阵乘积的形式，然后利用非负矩阵分解得到节点间的连接概率矩阵进行网络重建。设计实验并在几个真实的网络数据上测试，相比基于相似度的网络重构算法，该算法取得了更好的网络重构效果。%Based on the hypothesis that community structure determines the network connections , the connection probability matrix w hich describes the nodes ’ community structure can be trans‐fered into the form of product of matrices .T he nonnegative matrix factorization is applied here to get the connection probability matrix and then obtain the reconstruction .Experiments on several real world datasets show that the proposed algorithm outperforms some other algorithm which are based on similarity indexes .