A Strategy to Incorporate Voting Approach into Influential Nodes Identification of the Attention Flow Networks

Abstract

Identifying the influential nodes in complex networks plays a crucial role in the fields of epidemic control and public opinion guidance. Most of the previous algorithms are based on synthetic networks, the potential connection between nodes and higher-hop neighbors is not fully considered. Based on the massive online user behavior data, we construct a weighted collective attention flow network model with voting mechanism. By defining two indicators of voting contribution and voting score, the characteristics of high-hop neighbor nodes are effectively extracted, we finally convert the critical node identification task in complex networks into an influence maximization problem. Extensive experiments on five public datasets and private networks demonstrate that ANVM (Identification of influential node in attention flow network based on voting mechanism) algorithm can effectively avoid the problem of node influence overlap and the phenomenon of "rich club" in the network.