Improvement And Implementation Of Community Partition Algorithm Based On Weibo Following And Forwarding Relationship

Since the 21 st century, with the popularization of the Internet, human life has a delicate touch with network everywhere. People are not only a part of social network but also a part of the abstract nodes in the virtual network. With lots of smart phones put on the market and the era of mobile Internet arrival, demands of people fetching and sharing information to get response and communication become more faster than ever.People concern about interested information of other Weibo account or share some viewpoints by registering one for them. Everyone plays different roles on Weibo, they participate in or forward comments, and also concern the others’. Some roles are more important than the others, they’re key nodes of Weibo, their Weibo will bring great influence and interaction to other people, and so through the analysis of the key node, mining and utilization is the key research aim of this paper.Through carries on the analysis to the actual user data in Sina Weibo, mainly based on the interest of user community division, and then locked the corresponding key nodes in this paper. First proposed the relationship between community division use Fast Unfolding and Label Propagation based on Weibo. To eliminate the one-way attention caused by the noise, this paper screens the mutual concern relationship between users and then make community division to explore the corresponding key nodes, but the obtained experimental results are not ideal.After summarizing the analysis of the problem, according to the characteristics of Weibo, I found that the existence of another forward relationship of the forwarding behavior more reflected in the user’s interest. The dissemination of information is a dynamic process, pays close attention to the relationship is a process of relatively static, which in a long period of time will not be changed greatly, and with respect to the dynamic forwarding behavior can be a good show information flow process. Therefore,the paper proposes a method based on Weibo forwarding the relationship between community division. And then use the following two methods to make specific analysis:1. Use Fast Unfolding to divide the data of the community based on the forwarding relationship directly. To combine with the feature of small world and the data filtering, the results contain a large number of scripts account, which interfere with the results.To divide community right after by designing the formula of authority downgrading to drop jokes account down based on forwarding relationship, and then filtering only the amount of a-day interaction Weibo data, this method can exclude the interference script account accurately.2. Use Label Propagation to make the edge weighted based on forwarding relationship to let the Label Propagation algorithm vectoring. By weighting to edges between nodes, combined with health principle to partition dataset, finally get the health community and the corresponding key nodes.Thus through a study about user forward relationship of digging out many important hidden nodes of the network, it is a very important significant to us to understand and master, even guide the flow of forwarding information. It also have a very positive impact for advertising or information dissemination and the exposure, on one hand, a good put in effect will be to attract businesses to inject capital, to advertise the Weibo platform, on the other hand information communicated widely will enhance the Weibo influence. This method is applied and verify in the nation’s largest information sharing platform.