An optimised deep learning framework for fake news detection by leveraging social media and news content features

The proliferation of fake news on social media and news platforms threatens information integrity, necessitating advanced fake news detection techniques. In this context, the paper proposes an optimised deep learning framework for fake news detection that leverages features from social media and news content. In the proposed approach, GPT-2 is employed for feature extraction, and the Sand Cat Swarm Optimisation (SCSO) algorithm is used to fine-tune it. The extracted features are fed into a deep learning model, whose hyperparameters are optimised using the Marine Predators Algorithm (MPA). The proposed model achieves 0.79 accuracy, 0.7933 precision, 0.78831 recall, and 0.78706 F1-score.