论文部分内容阅读
Retweet behavior is the key mechanism for information diffusion on microblogging networks.Mining and understanding the latent mechanism of retweet behavior is important for predicting,controlling information propagation.In this paper,we firstly do some empirical analysis on both user,weibo,interaction and user-centered transfer networks.Base on the result of correlation,we choose user attributes,microblog content,interactive attributes and local structures as main features for predicting users retweet behavior.Comparing with other studies,our method takes into account not only users recent situation,such as users retweet activity,interactive strength and interests,but also users near neighbors influence and the diversity of local structures.Then,using these multifeatures,we utilize some different fashion supervised classifiers to predict retweet behavior on Sina Weibo dataset.The experiments and evaluation show the effectiveness of our feature choices.And the results show that the features we selected combined with the Logistic Regression model for predicting users retweet behavior more accurately.