Graph Neural Network based Approach for Rumor Detection on Social Networks

Abstract

In today’s society, social media usage has resulted in several challenges, including the widespread dissemination of rumors - unverified or false information that can significantly impact public perception and decision-making. As a result, detecting and preventing the spread of rumors is essential. Recent scientific studies have proposed various solutions that use both machine learning and deep learning techniques to identify rumors. In this paper, we investigate an approach that employs graph neural networks to detect rumors. Specifically, we represent posts and their responses as graphs, which are then processed through a Graph Attention Network (GAT) layer. The resulting representations are fed into a dense neural network for classification. Our experiments on the PHEME dataset show that our approach achieves satisfactory performance in identifying rumors. This study also provides a promising avenue for future research in the field of rumor detection using graph neural networks.