Researching Abnormal Group Detection Of Microblog By Fusing Mutil-Feature

In the era of the development of the Internet,the social network has become more and more similar to the real communication circle,even has been unable to split the point,and this has gradually become the most important way of people's daily life.Social networking in the Internet Age,greatly shortens the distance in time and space of the reality of the circle,makes information can be quickly published and shared in every corner of the world.As people increasingly rely on social networking,the variety of behaviors and shared information in online social networks have an increasing impact on real-life work,family,social and privacy.As a large number of complex and chaotic information is accepted by people,it will subtly influence on people's decision-making and awareness,and even make people's economic life,family and social life have received the potential threat.With the rapid development of social networks,its influence is also becoming more and more wide,there are a lot of unsafe users emerge in the social network,especially in the micro-blog network,unsafe users refers to those users who spread and upload a large number of false and malicious information to in order to obtain economic benefits.In this paper,we understand the current status of the research on the abnormal account of the micro-blog field at home and abroad,we find that most of the current research is based on a single abnormal account characteristic,which cause the current recognition method can only identify the abnormal accounts generated in a single environment.However,more and more abnormal accounts began to imitate the normal behavior of the account,so that the current identification method is gradually ineffective.Therefore,how to deepen the research of characteristics of individual accounts,discover the behavior of new abnormal account characteristics in bulk,fastly and accurately identificate micro-blog network abnormal account is a problem need to be solved.Aiming at the problem of group abnormal account identification in online micro-blog network,this paper firstly designs a novel method to describe the group characteristics of the account.On the basis of collecting and analyzing a large number of Sina micro-blog account information,we proposed a group feature constructing method of integrating the attribute-based,the relationship-based and the time-varying-features-based accounts,and build group feature vectors for each account.Then,this paper designs a method of microblog anomaly population identification based on spectral clustering,which uses the vector space model to calculate the similarity between two accounts,and constructs the similarity matrix.Then,by using the spectral clustering method,the similarity matrix is clustered to identify micro-blogged abnormal groups.Experimental results on the real data set show that the method has high accuracy in accounting recognition of abnormal groups.On the basis of the above work,we propose a more optimize method from the time efficiency of view,which is based on the self organizing feature map neural network.This method,optimising the parameters the neural network of the self organizing feature map,uses Gaussian function to decrease the radius of the neighborhood,convery power function optimal learning rate,and tune output layer neuron number,so that it can achieve a.better efficiency of recognizing abnormal group.Experiments on the real data set show that the time efficiency of the abnormal group account is obviously improved by the method.