Research Of Community Detection Algorithms Based On Label Propagation

The past decade has witnessed a rapid development and change of the Internet. Numerous social networks including blogs, forums, microblogs and other social medias have been cropping up, drawing people together and empowering them with new forms of collaboration and communication. From a data mining perspective, social networks generally consist of huge number of users and relationships which gather together to form several communities. With the research of Internet technology, the technique of social community detection is also developed with the unpredictable speed. In our real life, the technique of social detection has been heavily used for a long time, for example,online marketplaces recommend products by tapping on crowd wisdom via user shopping and reviewing interactions, and political movements benefit from new forms of engagement and collective actions.A user spreads a rumor on Weibo, then the rumor creats an information diffusion mode of a kind of'virus type', how to find out the rumormonger in a short time, all of those require knowledge of community detection about social networks.This article firstly introduces community detection technique and its theoretical analysis, in-cluding technology of web crawler for data crawling, several traditional algorithms, such as GN algorithm, FN algorithm, KL algorithm and so on, some methods of community evalu-ation.Then the article has analysed label propagation algorithm thoroughly.Compared with many community detection algorithms, label propagation algorithm has the characteristic of simple computation and easy implementation so that it has been widely applied. But the ba-sic label propagation algorithm also has the problem of poor stability, aiming at the existing problem of poor stability caused by the random ranking of nodes, the article proposes one la-bel propagation algorithm based on node similarity(NSLPA) and the other label propagation algorithm based on key nodes(KNLPA) to solve the problem of poor stability, and identifies that both the label propagation algorithm based on node similarity(NSLPA) and the label propagation algorithm based on key nodes(KNLPA) improve the stability effectively by the theoretical analysis and experimental results. Compared with the basic label propagation algorithm, the label propagation algorithm based on node similarity(NSLPA) and the label propagation algorithm based on key nodes(KNLPA) both have been improved effectively.