Link prediction in multiplex social networks: An information transmission approach

Abstract

In recent years, link prediction in multiplex networks has attracted increasing interest of researchers. Multiplex social networks that model different types of social relationships between the same set of entities in separate layers are a special case of multiplex networks. However, most existing methods usually ignore that new links can also be formed through information transmission. Therefore, we propose a novel link prediction method that applies information transmission approach to multiplex social networks in this paper. To begin with, we define a new index and three new ways of information transmission in a multiplex network. In this regard, the similarities of potential links in the target layer are computed based on the total amount of information they transmit each other via fusing information from all layers. At last, the interlayer relevance method is used to weight all layers. To evaluate the prediction performance of the proposed method, extensive experiments are implemented on eight real-world multiplex networks, and the experimental results demonstrate that the proposed method significantly outperforms several competing state-of-the-art methods in most cases.