Weibo Friends Recommended Based On Topic Analysis And Social Circle Discovery

With the persistent development and innovation of Internet technology, the social software Microblog has created a new social interactive mode by the characteristics of briefness, convenient and free style. Microblog has also become an important platform for people to share, spread and receive information. Therefore, it is a big problem for such a vital social media platform that how to maintain people’s relationship in real life and how to expand their communities. According to the recommended missions from Microblog’s users, this paper is going to take a deeper research on four aspects:(1). Creating an active microblog topic excavation method for coalescing user behaviors and time window. Firstly, bring in four common behaviors of Microblog users: forwarding, commenting,@ others and participating in topical discussion. Then transit these four behaviors into the LDA mode as guideline information. Secondly, use the time window to divide the microblog text into time slice, so that establish an active microblog topic model based on the user’s behaviors and time window. Finally, compare to the experiments, it confirms that the topic model has effective outcomes for analyzing the time-dependent changes of topics and analyzing the changes of topic intensity. It has great supporting effects for the next research of community discovery.(2). Creating a community testing method based on the microblog topics and users’ following relationships. Creating a calculation method by considering the microblog (invisible) content (users topics) and social network topological structure (users’ following relationships) comprehensively. First, use the model that mentioned in the last paragraph to choose a user topic from the microblog data. Then make a similarity calculation by the users’ following relationships. Secondly, choose an.initial clustering center of a community and make a clustering to the user node, and combine the communities with a high degree of overlapping as well. Finally, check the communities. This experiment confirms that the community testing method cannot only find out unknown communities, but also can find out some new community topics. Compare to effects of the traditional community testing method, this method has higher improvement from it. It has terrific supporting effects for the next step of follower recommendation mission.(3). Creating a microblog friends recommendation method based on the topic analysis and community findings. Firstly, use the testing method based on the microblog topics and users’ following relationships that has mentioned in the last paragraph to test the users’ communities; secondly, in the tested communities, use the microblog topic model to extract the users’ topic distribution, and calculate the similarity among the users; Finally, create the microblog friends recommendation method based on topic analysis and community findings by connecting the influences from the authoritative users to microblog friends recommendation. The experimental results confirm that, the microblog friends recommendation method can have effective recommendations to microblog users.(4). Use the research achievements above, to design and realize a prototype system of microblog users’ recommendation. Connect to the microbolg topic model, the microblog users’ community testing method and the microblog friends recommendation method based on the topic analysis and the community findings, designing and building a prototype system of microblog friends recommendation has important supporting to the recommendation mission.