MBIC: A Novel Influence Propagation Model for Membership-Based Influence Maximization in Social Networks

• Main Authors: Gang Xie, Yongming Chen, Hongtao Zhang, Yuanan Liu
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Influence maximization; membership marketing; social network
• Online Access: https://ieeexplore.ieee.org/document/8735769/

Influence maximization is the problem of finding a certain amount of seed nodes that can trigger the largest expected number of remaining nodes under a pre-defined influence propagation model. However, most studies in this field apply online social network analysis methods mechanically and take commodity marketing in a broad sense as the research background, whose models are relatively simple, lacking characterization of specific business situation. Membership is a kind of classical marketing method. By selling membership, the company provides differentiated services to members and ordinary users, stimulating the former to consume. In this paper, we focus on the membership business model and study the membership-based influence maximization problem. First, due to the models used by predecessors failing in meeting the particularity of the membership, a novel influence propagation model membership-based influence cascade (MBIC) is proposed. According to the characteristics of membership, the MBIC model divides the influence propagation process into two stages, the influence stage and the reference stage. At the same time, the concepts of activity and intimacy are introduced to better model real social networks. Then, we propose the influence-reference rank (MBIC) algorithm that quantifies the ability of nodes at each stage in the MBIC to solve the membership-based influence maximization problem. Finally, experiments using real-world dynamic social networks with up to 1.4 million edges are conducted. The experimental results show that the influence-reference rank (IRR) achieves a better performance than several alternative algorithms under acceptable complexity conditions under MBIC model.