Data exchange similarity based on flow field for link prediction problem

Link prediction is widely used in many applications in complex networks and it is also an alternative tool for community detection. Based on many kinds of hypotheses, lots of efforts have been made to solve the link prediction problem and research has shown that some link prediction heuristics are successful because they effectively estimate the similarity between nodes in some latent space. In this paper, we start from the data exchange process that may happen in reality and propose the network data flow field hypothesis. Based on this hypothesis, we proposed three flow-based methods to estimate the similarity between nodes. We test the proposed methods on twelve real-world complex networks and experimental results show that on some specific networks, the proposed network data flow heuristics perform impressively better than typical prediction algorithms.