An Attention-Based Hypergraph Neural Networkfor Fake News Detection: Design, Analysis, andPerformance Evaluation

Abstract

The accelerating development of social media has
largely contributed to the spread of fake news, presenting
critical problems to social credibility and the stability of society.
Conventional detection methods usually do not reflect the
multifaceted association between users, posts, and news content.
The paper introduces an attention-based Hypergraph Neural
Network (HGNN) to detect fake news based on the UPFD
dataset. In contrast to traditional graph-based models, which
only describe pair-wise relationships, the proposed model uses
hypergraphs to describe higher-order interactions among many
entities. The attention mechanism is incorporated in the model
so that it can give more priority to various hyperedges, which
enhances the learning of the features and the classification. The
experimental findings prove that the proposed scheme is more
accurate, more precise, more recall, and has a higher F1-score
than the traditional machine learning models and regular
Graph Neural Networks (GNNs). A comprehensive analysis of
the model's efficacy, constraints, and potential future
applications is also included in this work.