Analysis of Fake News Detection using Graph Neural Network (GNN) and Deep Learning

The rise of fake news in social networks has become a significant concern, necessitating innovative strategies to combat misinformation. One such approach is the use of graph neural networks (GNNs) to trace and analyze false information dissemination. GNNs are specialized DL model capable of processing graph-structured data, making them ideal for capturing intricate dependencies and relationships in the spread of fake news. Proposed implementation starting with data preparation, where a carefully curated dataset is essential. The news stories are processed, tokenized, and transformed into a graph representation to better understand the relationships between words or entities. Graph construction enables the formulation of ML problems as graph classification tasks. Graph embedding, achieved through methods like word2vec or GNNs, plays a crucial role in representing each node as a low-dimensional vector, capturing its semantic significance within the network. This technique facilitates ML activities, such as node classification and link prediction. Combining graph embeddings and GNNs for classification allows for a more precise identification of true and false news, considering the structure and content of news propagation graphs. Finally, evaluating false news detection involves performance metrics such as accuracy. Proposed implementation obtained detection accuracy of 94%. Striking a balance between minimizing false positives and false negatives is crucial to combat misinformation effectively. By employing this approach, society can create a more trustworthy information environment, enabling individuals to make informed decisions and reducing the negative impact of fake news.