Research On Mining Algorithm Of Microblogging Social Interest Circle Based On Closeness And Influence

Social network is a kind of web service that spring up rapidly in recent years. It wasbased on the online friendships between users and made people can share interests andactivities freely with friends. Twitter and Sina Weibo are the represents ofsecond-generation social networking, i.e. microblogging service. With thedevelopment of mobile Internet, microblogging has become the most popularinformation dissemination platform, and has a huge influence and massive user socialdata.In academia and industry, social network analysis and mining has become a hotresearch area, community detection as an important aspect of social network analysisalso received a great deal of attention. However, most of the existing methods areoriented to the global network, the algorithm for given node’s ego-centric network isvery lack; Existing methods mainly consider the network link structure, but ignore thecloseness, influence, user interactions and other social features; Current methods aremostly single classification method, multi-classification algorithm that can discoveroverlapping intertwined communities is not yet mature.Based on the above problems, this paper proposed microbolgging social interest circlemining algorithm based on closeness and influence. The proposed algorithm usingmicroblogging users’ personal interactive data, adopted a step-by-step expansionstrategy, taking in to account the link structure, user intimacy and influence at thesame time, the main works are as follows:1. In user first level relation graph, this paper using K-clique-community theoryfinding the seeds which able to be form user’s multiple social interest circles. Thisstep mainly considered the network link structure to generate high-quality circle coreand initially identified the number of social interest circles.2. Formally defined the closeness of users and the normalized distance between nodeswith an existing circle. Then we designed a greedy algorithm, to extend every socialinterest circle based on closeness.3. This paper extended the PageRank algorithm using Weibo users’ social behavior tocompute user influence, based which we extending social interest circle again incentric user’s second relation graph. In addition, we also designed a social interest circle label algorithm. This algorithmusing users’ interest tag as input, mining circle members’ commonality andcombination with influence and TF-IDF’s thinking, to tag each circle automatically.In order to verify the correctness and effectiveness of the proposed algorithms, wedeveloped a Sina Weibo third-party web application, which can show the result of ouralgorithm intuitively and collect user feedback data simultaneously. The comparisonexperiments show that the proposed method has excellent performance.