SAFS Social Article Features Stacking Model for Fake News Detection

With the rapid development of social media, people can more easily and quickly produce, transmit, receive and share information, but this also provides channels for the widespread dissemination of fake news. Therefore, it is of great practical importance how to properly detect fake news and curb its continued spread. In this paper, an integrated machine learning model for fake news detection, the Social-Article Features-Stacking Model (SAFS), is proposed based on the work of our predecessors, using natural language processing and social network analysis methods. The model achieves good experimental results by multi-feature fusion of textual features and social network features of fake news. From the experimental results on the FakeNewsNet dataset, the SAFS model greatly improves the f1 value of fake news detection by 16.86% compared with the Social-Article Fusion Model (SAF), the original paper model. Finally, the paper explores the effectiveness of different features in detecting fake news.