Spammer detection on Sina Micro-Blog

The popularity of Sina Micro-Blog has made it a great target for spammers. The spreading spammers reduce the political, economic and social value of the Sina Micro-Blog and leads to a bad user experience, which attracts many researchers´ attention. Here we introduce the background of the spammer detection and present a study of spammer detection on Micro-Blog. We collected a dataset that includes 1,900 users and more than 340,000 tweets. And different from previous studies which mainly focus on shallow machine learning algorithms, we also used the deep learning model to build a classifier which can be used to discriminate spammers from legitimate users. However, different from deep learning´s good performance in other areas, the experimental results suggest that the deep learning model can´t learn more information from the user profiles and tweets than the Support Vector Machine although it outperform the Naïve Bayes. The experimental results suggests that we can build a classifier which can correctly discriminate most of spammers from the normal ones, and the deep learning model is not necessarily better than the shallow ones.