The Evaluation Of User Influence In Online Social Network Sites With Noise Existence

Since the era of Web 2.0,Online Social Network sites(OSNs)have developed rapidly,millions of individuals become to be the users of OSNs,for the OSNs have the features of ”openness”,”low threshold”,”timeliness”,”content simplicity” and ”terminal expandability”.OSNs have evolved to be the main platform for information dissemination to the public.More and more enterprises began to pay attention to control the influentials,to guide the dissemination of information in OSNs,so as to achieve the best communication effect.Because these influentials have large number of high-profile fans,high activity,once the message is forwarded by these influentials,the message will obtain the second,the third,· · · and the final transfer by his fans in a short period of time.Therefore,information explosion can be achieved by controlment of these influentials.User influence evaluation is the main method of mining influentials in social network.The openness,mobility and low threshold of social networks make it possible to quickly aggregate large amounts of users and information,but also make it include a large number of redundant,false,malicious noise information and noise users who broadcast noise information(such as the spammers).Besides,user relationship structure,interaction intensity,interactive content and other attributes are in a dynamic change.These factors make those influentials”hidden” in the noisy large-scale dynamic complex system,which presents a realistic challenge to the existing social network user impact evaluation method.At present,the common social network user influence evaluation method can be divided into three categories: network structure method,user behavior based method and information content based method.However,the existing user influence evaluation models: 1)only consider limited number of attributes and are short of systematic analysis;2)neglect the existence of noise;3)they do not provide appropriate model to evaluate the dynamical user influence.Therefore,the existing methods are mainly static methods of cross-section data in non-noisy environment,and there is a great application limitation in the user influence evaluation of social network under the noise environment.Motivated by the above scenarios,this paper focus on the evaluation of user influence in OSNs under the noise environment.The author discusses the influence factors of user influence from multiple dimensions angle,analyzes the propagation law of noise information,identifies the noise users,and measure the dynamic characteristics of different dimensions.The mathematical model is used to quantify the attributes of the users,the law of the behavior,the trend of the topic,and the dynamic characteristics.In the end,an accurate evaluation of user's Influence in a nosiy complex system is realized.In summary,the evaluation of user influence is the precondition of social applications.In this paper we proposed user influence measurement models to solve the difficulty of user influence evaluation on OSNs,which are of great value both to the extension of relevant researches and to practical opinion leaders' detection and pre-detection.