An experimental study of fake news detection on COVID-19 Twitter data and a conceptual framework for the early detection and prevention of fake news in social media

Social media with high information diffusion has become the dominant source of fake news with questionable veracity. The paper aims to provide reliable information through social media during emergencies. The study conducted an experimental analysis to detect fake news on social media, using different machine learning and deep learning techniques. It also addressed the reliability of social media data through sentiment analysis and statistical analysis for pattern detection. The experimental study revealed that support vector machine, logistic regression, random forest, and passive aggressive classifiers, long short-term memory, and bi-LSTM showed the best results. Besides, it was observed from the sentiment analysis that, there were more neutral than negative tweets. The paper contributed a conceptual model of hierarchical classification for early detection and prevention of fake news in social media by analysing the user, content and context, thereby providing reliable social media data, and facilitating decision-making without any ambiguity during disasters.