异常链路的识别与分析

Identification and Analysis of Anomalous Links

[目的/意义]链路预测的一个重要应用是对异常链路进行识别和分析。[方法/过程]文章首先讨论了异常间隙和异常链路,以及异常链路的识别方法。结合具体场景,研究发现,利用AA指标识别出来的异常链路,可以发现这些链路是用以构成微观基本单元三角形结构的;而利用Katz指标识别出来的异常链路,可以发现那些只存在唯一合作组合关系的情况。之后,讨论了异常链路与网络连通性的相关问题,以及利用网络的鲁棒性和脆弱性来评价连边重要性排序算法的基本思路。[结果/结论]这一思路对于异常链路也是适用的,异常链路在保持网络的连通性方面有着重要贡献。[局限]由于链路预测指标和算法非常多,文中方法仍需要在大量真实网络中进行验证完善。

[ Purpose/significance ] An important application of link prediction is to identify and analyze the anomalous link. [ Method/process ] First, this paper discusses anomalous gap, anomalous link, and identification methods of anomalous link. For the specific scene, the paper finds that the abnormal links identified by Adamie-Adar are used to form the triangular structure of the basic unit, while the abnormal links detected by Katz can show the exclusive co-authorship. Then the relationship between the anomalous link and the network connectivity is summarized and the basic idea of using the network＇ s robustness and vulnerability to evaluate the algorithm of link importance ranking is elaborated. [ Result/conclusion] This idea is also applicable to the anomalous link, which has significant contribution in maintaining the connectivity of the network. [ Limitations ] As the indicators and algorithms of link prediction are very large, the method in this paper still needs to be verified in a large number of real networks.