Research On Friend Recommendation System Based On Overlapping Community Discovery In Microblog

Microblog has become a very popular social platform for the information exchange and spread.The recommendation of friends can help users find the information they are interested in the massive information,and also can solve the problem of information overload in the network effectively.It is of great significance to meet user needs,and develop more diversified services.It also makes sense to formulate targeted products,disseminate information,and place advertisements.A two-phase method is adopted to achieve microblog friend recommendation in this dissertation.In the first phase,the community structure in the microblog system is discovered.It has been shown that communities are usually overlapping and hierarchical.The overlap and hierarchy of the microblog network community is considered in this dissertation and the improved EAGLE(Agglomerative Hierarchical Clustering based on Maximal Clique)algorithm is used to discover the overlapping community structure in the microblog based on the similarity of user tags.In the second phase,the LDA(Latent Dirichlet Allocation)algorithm is used to obtain the topic distribution of each user,and the similarity of topics between the target user is calculated to obtain the Top-N friend recommendation list of the target user with relationship between users.Java and Python language is used in this dissertation to implement a complete friend recommendation system,and a lot of experiments have been done on the actual microblog dataset to verify the availability of the system.The microblog network is the research object in this dissertation.The main research method is the complex network theory,community discovery algorithm,topic model similarity calculation.Data mining ideas is used to process and store data in the microblog system,modeling and implement algorithm,do experiment and analysis results.Machine learning methods are also used to the model training and classification,and this recommendation system proved to have better recommendation.