Selecting Opinion Leaders using Skyline Query (SOL-SQ): a new approach
Malek Chebil, Mohamed Anis Bach Tobji, Sayda Elmi, Rim Jallouli · International Journal of Computers and Applications · 2025
Online social networks engage millions of users, some of them emerge as opinion leaders influencing others. Identifying opinion leaders within a specific domain is critical given their pivotal role in shaping public opinion in several fields. The literature has explored several methods to detect opinion leaders according to different metrics. In this paper, we propose a new hybrid method called Selecting Opinion Leaders using Skyline Query (SOL-SQ). It aims to identify opinion leaders by analyzing both user interaction and content. The proposed method employs machine learning for topic-based analysis to assess specific context user's influence. It also uses statistical metrics to evaluate engagement, graph-based metrics for network centrality, and PageRank to measure popularity. SOL-SQ selects automatically the key opinion leaders through multi-dimensional metrics without weighting, and without choosing the Top-N parameter, i.e. the N best opinion leaders to select like in most existing methods. We evaluated SOL-SQ against existing methods, comparing average similarity and standard deviation in identifying opinion leaders. Our method outperformed others in all identified domain categories. Additionally, a subjective evaluation showed that tweets of the opinion leaders produced by SOL-SQ had the highest influence scores across all four domain categories, based on expert judgment and quartile analysis.