ECP-Rank: A novel vital node identifying mechanism combining PageRank with link prediction index

Abstract

The problem of identifying vital nodes in complex networks has been widely investigated during recent years. There have been various algorithms for this problem. PageRank is one of the most widely used algorithms. While during the process of PageRank, a random walker always selects next arriving node from its neighborhood randomly and uniformly. But in real world, this selection is more likely to have “tendentiousness”. In this paper, we propose a novel vertex centrality mechanism which takes this kind of “tendentiousness” into consideration. We will adopt “link prediction index” to assign a centrality score to an edge ei,j. During our new vertex centrality mechanism, the probability that a random walker staying at vertex vi moves to its neighbor vj is proportional to the centrality of the edge ei,j, thus we name this new vertex centrality mechanism as Edge-Centrality-Preferential Ranking mechanism, ECP-Rank centrality for short. We apply ECP-Rank centrality on a large number of data sets, where the new method’s better performance than traditional PageRank and some other classical centrality methods are verified.