Performance Analysis of GCN, GNN, and GAT Models with Differentiable Pooling for Detection of Fake News

In the era of social networks, the spread of fake news has become a critical issue that poses significant challenges to society. It is often difficult to distinguish between real news and fake news due to the sophisticated methods used to create misleading information. This paper presents a comprehensive approach to detecting fake news. The study conducts a systematic review of the performance of the Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), and Graph Neural Networks using DiffPool (GNN-DP) in Twitter data. It takes into account the limitations identified in the previous studies and fine-tunes the parameters to obtain better results. Performance evaluation of the proposed model was done using metrics such as accuracy, F1 score, and area under the curve (AUC). In the results obtained, it is found that GCN, with its dynamic attention mechanism, outperforms GAT and GNN-DP by effectively highlighting the most relevant information within the graph. Next, we discuss the practical implications of fake news detection and the GCN model. Finally, we present the future scope in this area for further research.