Identification Of Spammers In Social Network Based On Gaussian Mixture Model

Social network is an important branch in sociology study.The accumulation of user behavior data on social network platform brings about the rapid development of network scale.This platform shows people's social psychology,economic politics and many other human characteristics.A lot of researches are done on the basic of it such as advertising,decision making to improve commercial benefits.This advantage provides convenience for profit.Some users make use of it to attack others.They are called spammers.Most of the existing technologies mainly rely on the relationship between users.However,due to the development of intelligent recommendation mechanism,user's connection is not based on their real preference.The relationships between users are vague.Therefore,this paper proposes a scheme which does not rely on the complex relationships but users' multi-dimensional behavior data.By using a small number of labeled data and combining with semi-supervised training process,the optimal model for classification is obtained.Due to spammers emerge in an endless stream,we design a shielding strategy based on reinforcement learning,which can make the network recover in a short time.This paper first introduces spammers,the spread of spam and related work.Through the observation of data,we construct the initial model and optimize it,and verify the performance of the model.On the base of classification,according to the degree of spammers' destructive to the network,we design the shielding strategy which makes the network and each node have the skill of centralized and distributed intelligence.It guarantees the normal function of the network.