Multi-heterogeneous Social Network User Association Research

User matching in Multi-heterogeneous social networks refers to associating users according to the information published by different social networks and their network users,and discovering the same user has important value in the fields of interest recommendation,community discovery,special personnel monitoring,and can make the research of social networks further.This thesis aims at two typical representatives of Chinese social networks-Sina Weibo and Baidu Tieba,to conduct research on social network users.The specific research content includes the following three aspects:(1)According to the characteristics of Weibo and Tieba,construct the attribute model required for multi-heterogeneous social network association research.The attribute model is constructed based on the following for classes of user attributes: user background information,user name information,user's interests and user's network structure attributes.By combining a user's four classes of attribute together,this thesis put forward a method for user similarity comparison that can improve the accuracy of user matching.(2)Secondly,this thesis proposes a characteristic attribute association method for different attributes.To describe a user's background attribute,a user's gender,age,location and educational background information is used to calculate similarity.Based on the user name attribute method,this thesis selected seven attributes of user name length,user name special character,user name digital,user name character combination mode,user name change mode,user name special feature,and user name pinyin feature to calculate similarity.Based on user interest attribute,this thesis proposes the APW(All position weighted similarity)method to calculate the similarity degree of user published content.Compared with the traditional text similarity calculation method,the user interest similarity can be calculated effectively.Based on user network structure attribute,this thesis uses LDA(Latent Dirichlet Allocation)algorithm to calculate the similarity of interest of users' fans,follows and Tieba.This method can effectively classify interests,and calculate the similarity of user network structure by calculating the similarity of user interest classification.(3)Finally,together with a user's multiple attributes,and in order to increase the accuracy of the correlation results,the analytic hierarchy process and principal component analysis are used to calculate the similarity weights.Through series of experimental comparison,the high accuracy weight allocation strategy is proposed.This thesis designs a user similarity calculation model based on multi-attribute for multi-heterogeneous social network user association research,which realizes the function of association analysis on users on two social networks.The data set constructed by Sina Weibo and Baidu Tieba user data improves and enhances the association accuracy of multi-heterogeneous social network users.