An integrated model for detecting spam microblogs combining latent paragraph and user vector

With the development of Internet technology, there is an increasing number of people who start to communicate with each other via social network sites (SNSs), while the growth of the amount of spam information is also awful, which degrades the quality of user experience to a certain extent. In this paper, we assumed a new integrated model combining latent paragraph vectors trained via neural network and user vectors under social networks to detect spam microblogs based on Sina Microblog site. We verified and optimized through the SVM and Naive Bayes Classifier, and compared with models which only depend on user or textual features, we can achieve 46.14% and 9.81% promotion of performance via our new integrated model.