Search

Opinion Maximization Problem (OMP) targets at selecting a subset of influential initial nodes and eventually generating the maximum opinion spread. The current OMP methods mainly pay attention to the improvement of efficient algorithms, which hardly obtain high efficiency and stable accuracy in large-scale social networks. In this paper, we study the OMP with the community-based approach. To be specific, we first formulate the OMP and construct the weight-based opinion model to estimate the dynamic change of opinion value. In particular, to generate the influential individuals, we propose a Community-based Approach for the OMP (CAOM), including: community detection, selection of candidate nodes and generation of seed nodes. Then, to reduce the computational complexity effectively and distribute seed nodes into the reasonable communities, the significant communities are devised. Based on the one-hop measure and the potential nodes of each community, the candidate nodes are selected. For each community, we acquire the influence score based on its neighbors within community and beyond community. Finally, we develop the two-hop measure and Elimination of Overlapping Influence (EOI) to determine seed nodes from candidate nodes. Experimental results in ten social networks demonstrate that CAOM can accelerate the opinion spread with smaller running time compared with the baselines. [ABSTRACT FROM AUTHOR]