Analysis of Hoax News Propagation Patterns on Threads Using GCN-Based Structural Node Embedding
TL;DR
This study analyzes the propagation patterns of hoax news on the Threads social media platform using the Graph Convolutional Network (GCN) method and demonstrates that GCN effectively represents user relationships, identifies influential nodes, and provides a comprehensive understanding of hoax propagation patterns on the Threads platform.
Abstract
The dissemination of hoax news on social media has become a significant issue because it can influence public opinion and accelerate the spread of misinformation. Although various approaches have been proposed for hoax detection, most of them rely on textual analysis and do not adequately capture the structural relationships among users. Therefore, this study aims to analyze the propagation patterns of hoax news on the Threads social media platform using the Graph Convolutional Network (GCN) method. Data was collected by crawling posts from the Threads platform using hoax-related keywords and represented as a social graph in which nodes represent user accounts, mentions, and hashtags, while edges represent interactions among users. The constructed graph consists of 425 nodes and 697 edges. The proposed GCN model was evaluated based on degree centrality, betweenness centrality, clustering coefficient, and Spearman correlation analysis. The highest clustering coefficient obtained was 0.8046, while the highest Spearman correlation coefficient between the GCN score and degree centrality reached 0.6118. The results demonstrate that GCN effectively represents user relationships, identifies influential nodes, and provides a comprehensive understanding of hoax propagation patterns on the Threads platform.