Research And Implementation On Key Techniques Of Virtual User Profiles Inferring On Social Networks

As an exchange and sharing information virtual platform,social network platform,contains a large number of information,in which the user profiles information plays an important role in the personalized recommendation,precision marketing and public opinion guidance and so on.If social network platform couldn't obtain the user profiles accurately,then it's impossible to provide quality services for users.Therefore,it's indispensable to study user profiles.But,different researchers have different definitions of virtual user profiles in this field,which leads to the problem of structural confusion(unclear parent-child relationship),limited applicability of speculative methods and poor scalability of research results when studying profiles inferring problems.In view of these problems,this paper has carried on the thorough research to the virtual user profiles definition and the inferring method.Firstly,the expression model of virtual user inferring is studied for the problem of inconsistent definition for virtual user inferring.By analyzing the social network user data,the four profiles of the user are described.At the same time,a calculation model for virtual user profiles based on model classification and neighborhoodcommunity association update is proposed for the problem of missing and inaccurate profiles.The model can control the update operation according to the characteristics of different profiles,and improve the accuracy of profiles inference while saving time and cost.Then,the calculation model of virtual user profiles is applied to the specific profiles inferring tasks.In view of the difference of profile aggregation strength,we researched two profiles of the gender and occupation.In the case of gender inference,we focus on analyzing the characteristics of user usage habits,and the feature selection and weight calculation method based on dictionary are proposed.A classification algorithm based on naive Bayesian fusion is adopted to make up the effect of simple feature superposition.In the occupation inference,we improved the existing thematic feature selection method,designed a classification algorithm based on the updating mechanism and the influence factors in this algorithm are analyzed.Experiments show that the algorithm can effectively improve the classification effect when it is applied to occupation inference.Finally,based on the above study,a prototype system for user profiles inferring is designed.The system is used to enrich the expression model of user profiles by analyzing the information of the users to be published,link the relational data,obtain the signals from the neighbors and the community,and finally infer the user's profiles category,so as to enhance the personalized recommendation service quality.