Fake News Detection Using Weighted Fine-Tuned BERT and Sparse Recurrent Neural Network

Fake news refers to misinformation or false reports shared in the form of images, articles, or videos, disguised as real news to manipulate people’s opinions. Recently, fake news and rumors have spread extensively and rapidly around the world. This has led to the production and propagation of inaccurate news articles. Therefore, it is necessary to restrict the spread of fake information in the media to establish confidence globally. For this purpose, this research proposes Weighted Fine-tuned-Bidirectional Encoder Representations from Transformers-based Sparse Recurrent Neural Network (WFT-BERT-SRNN) for fake news detection through Deep Learning (DL). Data preprocessing is established using stop word removal, tokenization, and stemming to eliminate unwanted phrases or words. Then, WFT-BERT is employed for feature extraction, and finally, SRNN is employed to detect and classify fake news as real or fake. WFT-BERT-SRNN achieves a superior accuracy of 0.9847, 0.9724, 0.9624, and 0.9725 on the BuzzFeed, PolitiFact, Fakeddit, and Weibo datasets compared to existing techniques like DeepFake and image caption-based technique.