Fusing Multi-Features for Rumor tweets detection in Wechat

Recently, rumor detection has attracted lots of research interest because of the quick development of social networks and the widely spread of rumor. The exiting studies is mainly about the open social network such as Weibo and Facebook. However, there is little research about closed social networks, especially Wechat. As a famous app with more than one billion daily active user, the tweets in Wechat can attract lots of attention, which gives the spread of rumor tweets more chance. In this paper, we empirically study the tweets spread in Wechat. We first analyze the text information and users’ behavior of Wechat tweets, then we come up with several important features needed for rumor detection. Furthermore, we propose a new model which can capture the impact of both users’ behavior and text information. We conduct experiments using a large tweets dataset collected in Wechat, the result shows that our proposed method can effectively detect the rumor tweets.