A Novel Method for Mining Influential Nodes and Hidden Valuable Relationships in Generic Flow Networks

Identifying influential nodes is largely studied for its widely application in various fields. However, the previous studies usually concentrate on the undirected weighted networks or binary networks, neglecting the orientation information, which is vital for flow networks. Although some key nodes discovery algorithms based on the basic idea of PageRank are proposed to analyze nodes centrality in directed weighted networks, these methods can not be applied to the generic flow networks directly. At the same time, original research gives no answer to the exact interplay among nodes, which may help us to dig out hidden valuable relationships in the networks. In this paper, we put forward a novel algorithm to recognize the important nodes and invisible influential connections based on the notion of random walk. The total flow but not the direct flow in the networks should be considered to evaluate the impact of nodes. Finally, applying this method to a clickstream network, we find it do give a better rankings on websites than PageRank algorithm and helpful to excavate the websites´ hidden "visitors providers". This method provides a network flow analysis frame work and may be meaningful to many research fields like epidemic diffusion, traffic flow analysis, implied lexical meaning mining and so on.