Graph Convolutional-Based Deep Residual Modeling for Rumor Detection on Social Media

The popularity and development of social media have made it more convenient to spread rumors, has become especially important detect rumors in massive amounts information. Most the traditional rumor detection methods use content or propagation structure mine characteristics, ignoring fusion characteristics their interaction. Therefore, a novel method based on heterogeneous convolutional networks is proposed. First, this paper constructs map that combines both explore interaction during obtain representation. On basis, uses deep residual graph neural network construct information current model. Finally, Twitter15 Twitter16 datasets verify proposed method. Experimental results show higher accuracy compared