Research On Spam Filtering Technology In Social Network

Because of the large number of users, information disseminate fast, widerange, strong interaction, online social networks have become the biggest andactive social platform. Such as Sina micro-blog, WeChat circle of friends, GooglePlus and other social networks have become the important part of internet user’slives. Social networks bring convenience to people’s lives, while become the toolof advertisers and spamers. Therefore, filting the spam and detecting the zombieaccount in online social networks have become a hot research topic.In this paper, we take Sina Weibo as research platform. Using data miningand machine learning methods to classify and sort Weibo accounts, so as to detectthe zombie accounts and filter spam effectively, the main work is divided into thefollowing four parts:At first, the online active learning algorithm of statistical machine learningis used to design and implement three kinds of online machine learningclassification systems. The three systems are based on logistic regression, naiveBayes, support vector machine, and used to classify the microblogs.Secondly, on the view of classification, because microblog belongs to shorttext, the features can be is few, feature vector is sparse, and the classificationeffect is limited, so Sequential Probability Ratio Test is applied to detect thezombie accounts based on the front microblog classifer result sequence.Moreover, for the unit to Weibo account, extract features from two aspectsof social behavior and microblog content, select the effective features and get theoffline model of libSVM to classify the Weibo account.Finally, on the view of sort, construct of the sub network social relationmatrix with the social relationship among the subnet accounts of Weibo platformand use PageRank algorithm to sort account value. Divide Weibo accounts togroups according to the ranking results, so as to determine spamers’ the scope. In one word, this paper extracts the features based on the content ofmicroblog and the social behavior of Weibo accounts. It can detect the zombieaccounts and reduce the spam in the social networks with the statistical machinelearning methods.