A novel group multi-criteria sorting approach integrating social network analysis for ability assessment of health rumor-refutation accounts

• Published in: Expert systems with applications Vol. 238; p. 121894
• Main Authors: Yin, Mengzi, Liu, Liyi, Cheng, Linqi, Li, Zongmin, Tu, Yan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.03.2024
• Subjects: Group multi-criteria sorting; Health rumor-refutation ability; Minimum discrimination information; Optimal clustering algorithm; Social media accounts; Social network analysis; Optimal clustering algorithm; Health rumor-refutation ability; Group multi-criteria sorting; Social network analysis; Minimum discrimination information; Social media accounts
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2023.121894

Blooming social media platforms provide breeding ground for health rumors. Despite the establishment of accounts by numerous organizations to counter health rumors, the effectiveness of these endeavors exhibits considerable variability. Thus, there exists a pressing need to refine the framework and operation of rumor-refutation accounts. Aiming at enhancing the proficiency of accounts in refuting health rumors on social media platforms and exploring the factors affecting it, this paper proposes a novel group multi-criteria sorting approach integrating social network analysis (SNA) to classify accounts’ health rumor-refutation ability. To commence, an evaluation indicator system for accounts’ health rumor-refutation ability is established using SNA. Subsequently, the indicator values are computed, incorporating methods such as triangular fuzzy number (TFN), a lite bert (ALBERT) pre-trained language model, and PageRank. Furthermore, hesitant fuzzy linguistic term set (HFLTS) and triangular intuitionistic fuzzy number (TIFN) are used to determine the expert weights and indicator weights. After that, on the basis of original best worst method-sort (BWM-Sort), classification boundaries are discovered creatively using optimal clustering (OC), and minimum discrimination information (MDI) is adopted as the objective function for priority assignment. Consequently, an OC-MDI-BWM-Sort method is newly proposed which offers distinct advantages in computational efficiency, information integration, decision-making objectivity, and result effectiveness. Lastly, regarding to four cases of widely circulated rumors, health rumor-refutation ability of 35 accounts on Weibo platform is classified using the proposed method. The findings underscore that merely 8.57% of accounts exhibit stable and good health rumor-refutation ability, while up to 28.57% and 80.00% display poor and inconsistent ability in certain instances. Tailored to accounts with excellent or good, satisfactory or fair, and poor health rumor-refutation ability, respectively, managerial suggestions are provided regarding information expression standards, account operator proficiency, and account cooperation, and all accounts are advised to watch audience behavior. •A health rumor-refutation accounts evaluation framework is developed.•Indicator system is built using SNA with values obtained by TFN, ALBERT and PageRank.•Expert weights and indicator weights are determined using HFLTS and TIFN.•BWM-Sort is optimized using machine learning-based OC algorithms and MDI.•Weibo accounts are classified and managerial suggestions are provided.