基于隐空间映射的顶点带属性网络的链接预测

Link prediction based on latent space mapping in networks with nodes attributes

对顶点带属性的网络中潜在的链接进行预测,提出一种基于隐空间映射的网络链接预测方法.首先将以邻接矩阵表示的拓扑空间和以属性信息表示的属性空间映射到同一个低维隐空间,使得映射后的隐空间尽量保持顶点间在拓扑空间和属性空间的相似性信息,然后根据顶点在隐空间的向量表达,计算顶点之间的相似度.结果表明,该方法预测精度高且运行时间短.

This paper presents a new latent space mapping based algorithm for link prediction on networks with nodes attributes.The algorithm maps the topological space and the attribute space to a lower dimensional latent space,so that the nodes’similarities in the original spaces are reserved in the latent space.Then similarity between nodes can be obtained by calculating the distances between the nodes in the latent space.Such similarity can reflect the potential characteristics of topology information and attribute information.This experimental results show that it can obtain higher quality results on networks with nodes attributes than other algorithms.