An efficient rumor detection model based on deep learning and flower pollination algorithm

In spite of the growing popularity of social media (Twitter, Facebook, etc.) as a source of news and data, its unfiltered nature often facilitates the spread of rumors or pieces of information that cannot be validated at the time they are shared. It is possible for false or unconfirmed information to spread like wildfire on the internet, influencing public opinion and policy in the same way that reliable news would. Some of the most pervasive examples of incorrect and dubious information are fake news and rumors, both need to be identified as soon as possible to prevent potentially destabilizing outcomes such as loss of life, reputation, or financial loss. This paper presents a pioneering study that integrates the flower pollination algorithm (FPA) with convolutional neural networks (CNNs) for enhanced rumor detection on social media platforms. We develop and test a model that leverages FPA to optimize the architecture and hyperparameters of CNNs, which significantly improves the accuracy and efficiency of detecting rumors. Using data from Twitter, the proposed model achieves a benchmark accuracy of 91.24%, outperforming existing state-of-the-art approaches. The novelty of this research lies in the application of a nature-inspired optimization algorithm to automate the fine-tuning of deep learning models, addressing the challenges of manual parameter selection and model scalability in dynamic information environments. This study contributes to the fields of misinformation detection and machine learning by providing a robust framework for real-time, adaptable rumor analysis.