Competitive Influence Minimization in Multi-Group Social Networks: An Opinion-Based Solution

• Published in: IEEE transactions on network science and engineering Vol. 9; no. 4; pp. 2617 - 2630
• Main Authors: Li, Yuan, Zhu, Jianming, Jiao, Jianbin, Zhang, Qi
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.07.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Competitive influence diffusion; Fake news; False information; Greedy algorithms; Heuristic methods; Integrated circuit modeling; markov chain; Markov chains; Markov processes; Minimization; multi-group; online social networks; opinion transfer; Optimization; Social networking (online); Social networks; Steady-state
• ISSN: 2327-4697; 2334-329X
• DOI: 10.1109/TNSE.2022.3168042

Misinformation control has been a vibrant subject of research in online social networks (OSNs). With the diversity of OSNs, we observe that the emergence of group has also notably increased the exposure rate of misinformation. In the process of misinformation dissemination, the opinions adopted by an individual not only depend on the choices made by individuals' peers, but also highly depend on the individuals' own knowledge. However, we find that individuals' opinions are not consistent in the misinformation dissemination, and they are transferred dynamically. Motivated by these facts, we do some novel works on the problem of competitive influence minimization in multi-group OSNs. Firstly, we propose a novel dynamic competitive diffusion model. Secondly, a spontaneous mechanism and a contact mechanism are introduced to analyze users' opinion transfer processes, based on probabilistic discrete-time Markov chains. Thirdly, in order to make negative opinions be minimized, we use this model to work on a new Opinion Minimization (OM) problem. To quantitatively analyze this problem, a greedy algorithm of Equilibrium Opinion Minimization (EOM) is performed to select seed nodes. Furthermore, the experiments show that our proposed EOM algorithm outperforms the state of the arts, e.g., classical heuristic algorithms and local greedy algorithms of Influence Minimization in terms of minimizing opinion in steady-state opinion distribution.