Research On Micro-blog User Tag Recommendation Algorithm

Following the Web2.0 era,various social networking platforms(Sina,Twitter)are rapidly becoming popular all over the world.The number of users who share their opinions and obtain information on micro-blog is increasing day by day,Massive amounts of micro-blog data is also generated.In the rapid explosion of social data platform,micro-blogging personalized recommendations can help group users to obtain useful resource information.With the popularity of micro-blog social platforms,the mining of potential interests of micro-blog users is the focus of current research workers.Micro-blog platform provided users a function with mark their own personalized tags,Each registered user can identify themselves with the form of tags based on personal characteristics and preferences,This type of user tagging behavior allows online media to accurately capture their interests and hobbies,and Recommended interested products and information for users.However,there are many micro-blog users that do not add any tags or add few tags to themselves.In order to solve this problem,This thesis considers the user's social relations and blog content at the same time,using the topic model to extract the potential interest of micro-blog users,and make use of the relationship between friends and the mutual trust relationship between non-friends in micro-blog,Recommend personalized tags to users of micro-blogging platform.The Mainly work of this thesis is as below:(1)In the research of micro-blog user tag recommendation,Use the LDA topic model to extract the potential preferences of micro-blog users.Users often publish some messages related to themselves through the micro-blog platform,These blog messages can reflect the user's potential interest.This thesis use the topic model to deal with the user's blog messages,Tap and quantify the user's interest characteristics.(2)In the research of micro-blog user tag recommendation,we proposed a problem of Noise Link Exist In Friend Relationship.Users can follow each other on the micro-blog platform to exchange and recommend information to each other in time.However,there are some congregation relations arising from the bandwagon effect in such friend relationships formed by the followed behavior,That is,the noise link.Because of different friends have different influence on the target users,this thesis uses the KL Divergence to measure the user's preferences similarities with friends,the more similar preferences,the greater the impact on the target user,and thus reducing the impact of noise links(Congregation)on the target user,and eventually get the noise reduction relationship.(3)In the research of micro-blog user tag recommendation,we proposed a problem of Mutual Trust Relationship Exist In Non-friends.Comments?forwards?mentions is common forms of micro-blog user interaction in social platforms.The interaction between users reflects the user's interest in the topic.However,it does not mean that the more frequent of user interaction,the trust between them is higher,Whether we are active in the micro-blogging platform is an important factor to be considered.In this thesis,the user interaction frequency is constrained by the user's activity to define the mutual trust relationship among users,Can effectively avoid situation of cold start due to fewer or no friends.(4)In the research of micro-blog user tag recommendation,we proposed a Reconmmendation Algorithm of Micro-blog User-Tag Based on Noise Reduction Relation Regularization(BN3R-MUTR).Nonnegative factorization the user-tag matrix,and the noise reduction relationship is taken as a regularization term to constrain the low-dimensional user feature vector and punishing the distance between of two similar users.After noise reduction,the larger the relational value is,the greater the penalty is and the greater the impact on target users is.Based on this,the objective function is obtained,The model is optimized and constrained via the Lagrange multiplier method and the KKT conditions,and finally gets the approximate user-tag matrix for recommended users' tag.The proposed algorithm is validated using two different data sets.(5)In the research of micro-blog user tag recommendation,we proposed a Recommendation Algorithm of Micro-blog User-Tag Recommendation Algorithm Based on Social Relationship Regularization(BSRR-MUTR).In the social network micro-blog,friendship is important,but the interactive trust generated through dynamic interactions can not be ignored either.It can effectively avoid the cold-start problem caused by a small number of friends or no friends Using parameters to control noise reduction friend relationship and interactive trust relationship in the proportion of social relations,define the social relationship regularization term to constrain the user tag matrix and make the recommendation of the tag for users.