ERIUE: Evidential reasoning-based influential users evaluation in social networks

Abstract

Social media users are playing an increasingly important role in disseminating information, but their ability to diffuse information may vary significantly. Therefore, evaluating the influential ability of users has become crucial to promote or curb the dissemination of specific information. Existing centrality measures have produced varying results in identifying the most influential users, but it remains a challenge to identify the most influential users in a multifaceted and consistent way in social networks, especially when only a limited number of users can be nominated. To fill this gap, this work developed an evidential reasoning-based influential users evaluation (ERIUE) model that considers multiple sources of structural information in networks. Our proposed model collates information about users’ influential ability from multiple forms of centrality measures and maps their scores to different grades in an informative belief distribution. To determine the weight of each centrality, three types of information are considered: conflict of belief distributions, similarity of probability sets, and overlap of evaluations. The information is aggregated using the recursive evidential reasoning approach based on a formulated criterion hierarchy, thereby determining the influential ability of users. The applicability of our proposed model is demonstrated by comparing it with existing measures in three real-world social networks. Our proposed model is also applicable to relevant problems beyond identifying influential users, including preventing epidemic spread, cascade failure, and misinformation dissemination in social networks.