The Research And Implementation Of The Micro-blog Influence Evalution Model Based On Clustering Algorithm

Microblog is a new social networking tool, which sweeps around the world rapidlybecause of using conveniently and posting message real-time. Micro-blog influencereflects the user’s actual social influence.Evaluating of micro-blog influence accurately notonly can help users find valuable information,but also provides further expandingapplications for micro-blog.However,the existing models of micro-blog influenceevaluation has some disadvantages:firstly,a little experiment data lack ofconvincing;secondly,affected seriously by zombie fans;thirdly,can’t combine with theusers’ keywords;fourthly, low algorithm efficiency.Ordering to solve the existing problems of micro-blog influence evaluation models,this paper studied the following work:Firstly, improved the way of capturing data from Sina micro-blog, and increased theamount of experimental data. Sina micro-blog provides the API for developers to capturedata, but limits the API’s calls frequency. Through using multi-application and multi-agentIP technology to improve the method of fetching data from micro-blog API, fetched datafrom150times per hour increased to6000times per hour. Fetched1000000micro-blogusers,which increased the amount of data in the experimental model, and ensured theaccuracy of micro-blog influence evaluation.Secondly, put forward a method of discriminating and excluding micro-blog zombies,reduce the effects of zombie on evaluating micro-blog influence. According to themicro-blog’s huge users, analyzed in detail the difference between the latest zombies andordinary users, proposed an BP neural network Based on simulated annealing algorithmSAVBP,and realized a zombie classification system based on SAVBP. Used Sinamicro-blog data to evaluate the system, results showed that the system increased thediscriminating accuracy rate and recall rate.Thirdly,proposed a micro-blog influence evaluation model that based on clusteringalgorithm.firstly preprocess the data, exclude micro-blog zombies.Established a complexnetwork based on the attention link between micro-blog,and clustered the network by aimproved Girvan-Newman algorithm. Decomposed the cluster structure and to thisstructure proposing a micro-blog influence evaluation model based on user influencekeywords CRank. Used Sina micro-blog data to evaluate the model,the result showed that the model has good convergence, computational efficiency is superior to the traditionalmicro-blog influence evaluation model based on PageRank algorithm.Fourthly,designed and implemented the micro-blog influence evaluation model basedclustering algorithm.There is no uniform standard for the evaluation of the influence ofmicro-blog evaluation model, this paper put the customer satisfaction as the evaluationcriteria micro-blog influence evaluation model. Compared with the existing micro-bloginfluence evaluation model based on Hits algorithm and based on PageRank algorithm, theresults showed micro-blogging influence evaluation model based on clustering algorithmhas a high real-time and user satisfaction.With the rapid development of micro-blog, the evaluating of micro-blog influencewill become increasingly important, influence evaluation model of this paper provides afoundation to further expanding of the micro-blog application.