Research on the influence maximization based on community detection

Abstract

The research of influence maximization aims to select the k most influential nodes in social network, these nodes are a group of initial seed nodes for information dissemination and make the ultimate scope of influence maximum. Considering that Community-based greedy algorithm has two problems. One is that community division is unstable, and the other is that the time complexity of selecting seed nodes on the basis of community division is too high, this paper proposes an influence maximization algorithm (LPIMA) based on community detection. The algorithm selects optimized label propagation algorithm for community detection. Firstly, we use LeaderRank algorithm to quantify the influence of community nodes. And then assign candidate seed node set according to these quantified values. Finally, we use the submodel characteristic to improve greedy algorithm and mining community seed nodes from candidate seed sets. The results show that the proposed algorithm guarantees the scope of influence and improves the time efficiency in the large-scale network.