Fake News Detection Using Hybrid Transformer-Based Model

The rapid spread of fake news on digital platforms threatens public trust and social stability. This paper proposes a hybrid deep learning model combining Robustly Optimized BERT Pretraining Approach (RoBERTa), Graph Neural Networks (GNN), and Heterogeneous Attention Networks (HAN) to improve fake news detection. Existing models capture contextual information but struggle with complex entity relationships and hierarchical data structures. Our hybrid approach leverages RoBERTa’s robust language understanding, GNN’s relational modelling , and HAN’s hierarchical attention to address these limitations. The model is evaluated through classification, prediction, and baseline comparison modules, using accuracy, precision, recall, and F1-score metrics. Experimental results demonstrate that the proposed model achieves an outstanding 99.77% across all these metrics, significantly outperforming traditional and baseline methods and providing a highly effective solution for fake news detection.