Rumor detection on social media through mining the social circles with high homogeneity.

The massive spread of rumors on social media has become a major global challenge, increasing the urgent demand for rumor detection. Although social circles are ubiquitous in social networks and have the property of describing users' behavioral preferences, they have not been explicitly considered in rumor detection models. Meanwhile, information diffusion studies have shown that social circles have a significant impact on the speed, scope, and content of rumor propagation. Motivated by this important absence, we validate the significant difference between the social circles of rumor and non-rumor sources, and propose a new rumor detection algorithm. The algorithm explores a new feature space by extracting social circles with high homogeneity from user context, and combines it with social interaction to automatically detect rumors. Experimental results obtained on three real-world datasets support that the proposed approach outperforms state-of-the-art methods and displays a superior capacity for detecting rumors at early stages. The code of this work is made publicly available to foster any further research.1