Predicting user retweet behaviors is the basis of building the information diffusion model in micorblog social networks

and also is applied for a variety of fields such as public opinion monitoring

viral marketing

political campaign etc.In order to improve the accuracy of predicting user retweet behaviors

under the MRF (Markov Random Field) framework

the paper comprehensively analyzes the effects caused by user attributes

microblog contents

the constraints between user retweet behaviors and the group retweet priors

and constructs the corresponding energy function based on the logistic regression model to globally predict user retweet behaviors.Experimental results show user retweet behaviors not only depend on user attributes

and micorblog contents

but also are influenced by the constraints between user retweet behaviors and the group retweet priors in varying degrees.Compared to the traditional methods

our proposed method can accurately model user retweet behaviors and thus achieve satisfactory results.