基于拓扑信息耦合加权的链路预测方法

Link Prediction Method Based on Information Coupling of Topology Weighting

链路预测方法的提出往往需要一些好的网络生长机制来支持,如网络的社团结构、偏好连接和弱连接效应都可以很好的指导链路预测.现有的基于相似性的方法,忽略了拓扑信息耦合对实际网络演化的促进作用.鉴于此,提出一种基于拓扑信息耦合加权的预测方法.首先,以拓扑信息耦合度构建边权矩阵,进而为节点间不同异构的路径计算其耦合度;其次,综合多跳不同长度路径对相似性的贡献评价节点间的相似度.实验结果表明,提出的指标具有较高的预测精度,且具有良好的鲁棒性.

The proposal of link prediction methods often requires some good network growth mechanisms to support,such as the network community structure,preference connection and weak connection effects can all guide link prediction very well.The existing similarity-based methods neglect the promotion of information coupling of topology to real network evolution.In order to solve this problem,this paper proposes a prediction method based on information coupling of topology weighting.Initially the link weight matrix is constructed by the coupling degree of topology information and the coupling degree for different heterogeneous paths between nodes is calculated Then,the similarity between nodes is evaluated by integrating the contribution of multi-top paths with different lengths to the similarity.Empirical study shows that the proposed index has high prediction accuracy and good robustness.