A Deep Learning Model for Fake News Detection Using BERT and Ensemble Techniques

The rampant dissemination of fake news across digital platforms has emerged as a profound challenge, undermining public trust, democratic processes, and health-related decision-making. This study proposes a robust deep learning model, incorporating Bidirectional Encoder Representations from Transformers (BERT) and ensemble techniques for the detection of fake news within healthcare and political domains. By leveraging domain-specific datasets (CoAID and PolitiFact), and integrating both traditional (TF-IDF, n-grams) and advanced) features, the model demonstrates superior capability in discerning authentic from fabricated content. The proposed system significantly outperforms baseline classifiers, with BERT-based models achieving high accuracy and F1 scores. This research also underscores the role of psychological, social, and technological factors in the propagation of misinformation, and advocates for scalable, domain-sensitive, AI-powered solutions for real-time fake news detection.