Spammer Detection And Its Application Research In Marketing Information Dissemination Model On Social Network

With the continuous development and penetration of social network,it has owned hundreds of millions of users.Enterprises often take advantages of such large-scale users to promote products,maintain brands,and build customer relationships on social network.However,there exists a kind of users,namely spammers.Driven by interests,they purposefully produce larger-scale false behaviors such as praise,comment and forward,in order to guide public opinions,promote products and control word-of-mouths.Some even vilify competitors and related products.The existence of spammer not only seriously affects the fair marketing environment of social network,but also some even violates business morality,which can lead to extremely bad influence.Traditional research focuses on the user static behavior features to detect spammers.As strategy of spammers evolves,spammers become more and more close to normal users,such method has been unable to effectively detect increasingly hidden spammers.Therefore,it is increasingly important to observe and analyze the dynamic behavior features of spammers.Moreover,traditional model detects spammers mostly only in offline environment.Due to the persistence of social network data,traditional model is difficult to apply and cannot adapt to the environment of online detection of social network.Spammers generate a large number of purposeful behaviors on social network,which interfere with the process of marketing information dissemination and disrupt the fair marketing environment of social network.However,in the work of marketing information dissemination model of social network,predecessors have less considered the influencing factor of spammer in a fine-grained manner.Only by researching the impact of spammer on the process of marketing information dissemination in a fine-grained manner,finding the internal mechanism of the spammers,can we effectively govern and deal with the harm brought by spammers and maintain a fair and orderly social network marketing environment.The main work and innovation of this paper are as follows:1)Because spammers strategy is becoming more constantly changing and concealed,and traditional model is difficult to apply to the online detection environment of social network,this paper used calculating-volatility method to construct the dynamic behavior features of the social network users.It has low time complexity and convenient for practical use;the difference and effectiveness of the dynamic behavior features between spammers and normal users were compared and verified in offline environment;then the online spammer detection model was constructed by combining the user's dynamic and static behavior features with the semi-supervised model Tri-Training.Therefore,this paper not only extended the study of behavior features in spammer detection field: from static behavior features to dynamic behavior features,but also made up for the research deficiency in online detection environment.2)In view of the problem that spammers will interfere with the process of marketing information dissemination and disturb the social network marketing environment,there are few works to take into account the influence of the spammer in a fine-grained manner in the marketing information dissemination model of social network.So,based on the Agent-Based IC model,this paper constructed a marketing information dissemination model with social network marketing theory;a marketing information dissemination model considering spammer was also constructed by incorporating influencing factor of the spammer into the model.3)Finally,by crawling the user data of Sina Weibo,the differences and effectiveness of the user's dynamic behavior features were compared and analyzed in the offline environment,and the effectiveness of the online spammer detection model was verified: the F1 value is 93.33%,average training time is about 2 minutes;by crawling the marketing information data of Sina Weibo,the effectiveness of the marketing information dissemination model considering spammer was verified: the heat and exposure are more closer with the actual dissemination results,in addition,the impact of spammer on the process of marketing information dissemination was analyzed and relevant governance advice for individuals,enterprises and relevant departments was provided with the parameters of the model.