从信息学的角度分析复杂网络链路预测

链路预测作为复杂网络分析的一项重要任务,其目的是寻找节点间缺失(新)的链路,识别虚假交互,对于挖掘和分析网络的演化,重塑网络模型具有重要意义.传统的链路预测方法多数采用拓扑结构信息、节点的属性信息和图的结构特征.应用这些特征等外部信息可以得到很好的预测效果.本文从信息学的角度全面分析、回顾和讨论了复杂网络链路预测的发展现状,提出了链路预测技术和问题的系统分类.首次将分层的思想引入链路预测分类体系中,把当前的链路预测方法分为基于监督学习的技术、基于半监督学习的技术、基于无监督学习的技术和基于强化学习的技术.对每种技术的优缺点、复杂性、所使用的具体特征,开源实现及应用建议进行了详细的分析.最后,讨论了当前复杂网络链路预测技术未来的发展方向.

Link Prediction of Complex Networks is Analyzed from the Perspective of Informatics

As an important task of complex network analysis,link prediction aims to find missing( new) links between nodes and identify false interactions,which is of great significance for mining and analyzing the evolution of networks and reshaping network models.The traditional link prediction methods mostly adopt topology structure information,node attribute information and graph structure features.Using these features and other external information can obtain a good prediction effect. This article from the perspective of informatics comprehensive analysis,reviewed and discussed the current situation of the development of complex network link prediction,classification of link prediction technology and problems of the system was put forward. This paper analyzes,reviews and discusses the development status of complex network link prediction comprehensively from the perspective of informatics,and puts forward the systematic classification of link prediction technology and problems. For the first time,the hierarchical idea is introduced into the classification system of link prediction,the current link prediction methods are divided into supervised learning technology,semi-supervised learning technology,unsupervised learning technology and reinforcement learning technology. The advantages and disadvantages of each technology,complexity,specific features used,and open source implementation and application recommendations are analyzed in detain. Finally,the future development direction of current complex network link prediction technology is discussed.