融合节点信息和社区信息的复杂网络链路预测

Research on Complex Network Link Prediction Combining Node Similarity and Community Information

链路预测作为复杂网络中一个充满挑战的研究方向,具有非常广泛的应用前景。链路预测被称为对网络中缺失或未观察到的链接的预测,最前沿的链路预测方法要么仅考虑节点之间相似性,要么仅简单挖掘社区之间的信息,并没有达到很好预测目的。为了解决上述问题,论文提出了一种衡量重叠社区与非重叠社区关系强度的度量标准,且为了更好地考虑社区之间的信息对预测的影响,还引入了新的社区划分方法,最后提出了一种同时考虑节点相似性和社区结构信息的链接预测框架。实验结果表明,与目前已有的算法相比,论文提出的链路预测算法在AUC精度上提升了0.2%~10.59%,证明了论文提出的方法是有效的。

As a challenging research direction in complex networks,link prediction has a very broad application prospect.Link prediction is called the prediction of missing or unobserved links in the network.The most cutting-edge link prediction meth-ods either only consider the similarity between nodes or simply mine the information between communities,and they don't achieve very good forecast purposes.In order to solve the above problems,this paper proposes a metric to measure the strength of the rela-tionship between overlapping communities and non-overlapping communities,and in order to better consider the impact of informa-tion between communities on predictions,a new community division method is also introduced,and finally a link prediction frame-work that considers node similarity and community structure information is proposed at the same time.Experimental results show that compared with the existing algorithms,the link prediction algorithm proposed in this paper improves the AUC accuracy by 0.2~10.59 percentage points,which proves that the method proposed in this paper is effective.