复杂网络高阶结构的关联规则挖掘及其应用

Association rule mining and applications based on higher-order structures in complex networks

  • 摘要: 网络高阶结构即满足特定条件的子网络,是网络科学领域重要的研究内容。近年来,关于高阶结构的研究不断增加,但是关于高阶结构之间内在联系的研究还相对较少。基于此,根据传统关联规则,定义了高阶结构之间的关联规则评判指标,并提出了一种有效挖掘高阶结构之间关联规则的通用算法框架。利用该算法,在6个真实世界网络中进行了3阶高阶结构(即高阶结构包含3个结点)间的关联规则挖掘。实验结果表明,真实世界网络中高阶结构之间存在强关联规则,且不同真实世界网络中高阶结构之间的关联规则存在差异。此外,将挖掘出的强关联规则应用于链路预测当中,进而实现了链路预测方法。相比于基线方法,所实现的链路预测方法在4个真实世界网络中取得了最好的性能表现。

     

    Abstract: The study of higher-order structures, which refer to subnetworks within a network, is a crucial research topic in network science. In recent years, although the research on higher-order structures has been increasing, there has been relatively little research on the internal connections between higher-order structures. In light of conventional association rules, the evaluation criteria of association rules between higher-order structures are defined, and a general algorithm framework for effectively mining these association rules is proposed. The proposed approach has been applyed to mine association rules among three-order structures in six real-world networks. The results demonstrate strong association rules between higher-order structures in real-world networks, as well as variations in these rules across different networks. Additionally, we apply strong association rules to link prediction, resulting in a new link prediction method. This method outperforms the baseline methods in four real networks.

     

/

返回文章
返回