Research On Microblog Recommendation Method Based On Weighted Dynamic Interestingness

Microblog is a social networking platform for users to share information in real time by focusing on relationships. Different users may have the same preference, and then come into being community network. This gives the precise positioning of the user's interest orientation, the organization provides accurate information to provide the possibility of publishing, increasing the probability of users get their own interest information. Therefore, the user's interest has become a hot research topic since the emergence of microblog.There are a lot of researches on microblog data mining analysis. But most of these studies focus on the structure of social networks or the analysis of microblog's text. Although there are many models of microblog about user interest, but almost no one can systematically explore the user interest in the perfect microblog interest model. Because the user's interest will change over the time, therefore this paper introduces the time factor. Firstly, according to the existing latent Dirichlet distribution model, we calculate the subject distribution of the data set of microblog. Secondly, according to the frequency and similarity of the interaction between the user and the user of microblog,thus we can get the interest degree of user's interest aggregation.Furthermore, weighted the user's interests and the set of user's interests,and we can get more accurate interest degree for microblog user to the theme. Finally, through a new microblog theme distribution, as well as the interest degree of the new microblog users to the theme of microblog.Thus, the weighted dynamic interest degree can be concluded. At this point, calculated the user's weighted dynamic interest degree to the new microblog one by one. And then arranged in descending order,according to the degree of interest reduction algorithm. Then, Top-N microblog can be recommended to the user, thereby achieve accurate delivery.We analyze the proposed model in this paper from the aspects of overall accuracy, time accuracy and the influence of different weights on the model. And compared the proposed model with the collaborative filtering algorithm based on LDA model and RT-LDA model. The results of comparative analysis show that in this paper, we propose a recommendation model which can be used to reflect the user's interest more accurately than the traditional model.