Information spreading with relative attributes on signed networks

Abstract During the past years, network dynamics has been widely investigated in various disciplines. As a practical and convenient description for social networks, signed networks have also garnered significant attention. In this work, we study information spreading with relative attributes on signed networks, where edges are assigned positive or negative labels, describing friendly or hostile relationships. We define the attribute of information by a degree that can be either ‘good’ or ‘bad’ and assume that the spreading willingness of the information receiver depends on not only its relation with others but also the attribute of information. A pair-wise potential relation identification algorithm is designed based on the shortest path approach and structural balance theory. Both simulations on randomly signed networks and empirical experiments on real datasets show that the proposed information spreading could be approximately investigated within a local 2-order neighborhood. In addition, the ratio of potential friendly nodes with a target node is consist with network content. Finally, the propagation speed of ‘good’ information would unexpectedly slow down when the ratio of positive edges is larger than an estimated threshold. The presented model could be referred to in real social scenarios, such as product promotion, advertisement media, and rumor mongering.