Identifying influential nodes in social networks: A voting approach

With the prosperity of social networks, the research of influence maximization has become growing importance and captured increasing attention from various disciplines. The key point in influence maximization is to identify a group of influential nodes that are scattered broadly in a network. In this regard, we propose the VoteRank++ method, which is a voting approach, to iteratively select the influential nodes. In the viewpoint of VoteRank++, nodes with different degrees should carry different amounts of votes in consideration of the diversity of nodes in voting ability, and a node may vote differently for its neighbors by considering the varying degrees of closeness between nodes. Moreover, to reduce the overlapping of influential regions of spreaders, VoteRank++ discounts the voting ability of 2-hop neighbors of the selected nodes. Then, to avoid the cost of calculating the voting scores of all nodes in each iteration, only the nodes whose scores may change need to update their voting scores. To demonstrate the effectiveness of the proposed method, we employ both the Susceptible-Infected-Recovered and Linear Threshold models to simulate the spreading progress. Experimental results show that VoteRank++ outperforms the baselines on both spreading speed and infected scale in most of the cases.