Research On User Interest Modeling Based On Microblog Social Network

Micro-Blog is an information-sharing, disseminating and accessingplatform based on user relationship. It’s a newly raised open Internetsocial service which is mostly known as web2.0. In a Micro-Blog system,relationship network is formed by follow or followed by each other. Theinformation user published broadcast in an amazing speed through thefunction called forward. In this situation Micro-Blog has become not onlya social network but also an important media which provide large amountof fresh news and comments. Recently more and more foreign companiesstart to research on user interest in Micro-Blog because of the commercialpotential behind the market. There is no such system research in domesticcurrently, so this paper proposes the following methods and ideas for userinterest modeling:Firstly, go through the information structure of Micro-Blog socialnetwork, summarizes the composition of user interest. Design aMicro-Blog information collection system which is more efficient thantraditional collection program. The distributed collection system performsmore efficiency and scalability in practical applications than thetraditional programs;Secondly, build a real Micro-Blog social network. A new algorithmis proposed based on the degree of convergence algorithm Micro-Bloguser-important degree combined with complex network node importantdegree algorithm theory, which reflect the Micro-Blog users realityinfluence;Thirdly, do research on the Micro-Blog short text keywordextraction including the redundant information filtering. Construct graphaccording to the similarity of the relationship between the words bynatural language processing segmentation tools. Through the adjacency matrix calculate the degree of importance of the word, thereby extractingthe key words in the text of the micro-Blog;Fourthly, on the basis of user importance and Micro-Blog Keywordsproposed feature item extraction algorithm based on improved TF-IDFalgorithm. Build user interest space by vector space model, throughlong-term and short-term interest get the definition of the user’s degree ofinterest, complete user interest modeling. Demonstrate the model hasgood accuracy compared with artificial works by experiments.