Detection of fake news in social media using CNN with grey wolf optimized BERT

Fake news refers to stories that are falsely represented as news, that has the potential to misinform and mislead readers and identifying such fake news is a significant challenge to combat its impact in society. Our proposed work detects fake news using a hybrid method for fake news detection that uses Bidirectional Encoder Representations from Transformers (BERT) framework for feature extraction, Convolutional Neural Networks (CNN) for classification and Grey Wolf optimization for optimizing the parameters of BERT. The raw data is preprocessed and transformed into a format that can be fed as an input to the BERT model for feature extraction. This pre-trained model is fine-tuned using Grey wolf optimization (GWO) and CNN is used to classify each news article as real or fake. The proposed BERT-GWO-CNN model outperforms the existing machine learning models like Random Forest, SVM and deep learning namely RNN and LSTM in terms of prediction accuracy.