Identifying spammers in Microblog websites is an interesting research direction Most recentlv methods of identifying spammers exploit user attributes such as their tweets.registration time.number of friends.and so on These methods perform poorly on a real dataset In this paper we thoroughly compared and analvzed various user attributes and behaviors of common users and spammers As a result.user profile features and behavior features are defined to distinguish common users and spammers We propose a probabilistic graphical model named SPGM (Spammers ProbabilisticGraphical Model to detecting spammers of Microblog websites In this SPGM model.we treat user profile features as the input variables while user behavior features act as the observed variables and the probabilitv of a user being a spammer is a hidden variable of SPGM model After collecting data from Sina Weibo and Twitter respectively.we build two experimental datasets to evaluate the performance of SPGM model In our experiments.we choose different user features for these two datasets and compare the learning ability and the predictive abilitv of our SPGM model running on different experimental two datasets We also compare our SPGM model with the baseline SVM model The experimental results prove that the SPGM model can effectively detect spammers and predict new spammers in the Sina Weibo dataset and Twitter dataset SPGM model outperforms SVM model on detecting Microblog spammers.