User Interest Analysis And Prediction Based On Microblog

Microblog, as a new way of information sharing and dissemination, has been especially popular in recent years. The large number of microblog users and rich contents make microblog not only to become a way to access information for users, but also be a highly effective advertising platform. In the vast microblog information, users will take a lot of time and effort to find the information they interested. So the information recommendation service appears. Mining user’s interest is the core technology for precision advertising and information recommendation service.After researching the related works, this paper proposes to analyze the user’s interest on their microblog data sets, and further to predict the users’ recent changes in their interest. First of all, for a given user, it collects the user’s microblog data sets and filters them using the pre-built filter. Secondly, on the external knowledge base to build theme model first, and then apply the model to analyze the users’ microblog contents. Thirdly, the user’s interest is represented by the topic probability combined with the time weight. Finally, this paper analyzes the microblog data sets of friends which user followed using the maximum entropy model, and mines the user’s interest recently for a period of time, so as to achieve the purpose of forecasting.This approach has the following advantages. First, considering the microblog text is very short, we propose to train the topic model combining the external knowledge in order to expand the microblog content semantically. Second, when expressed the user’s interest, we coupled with a time weight for each topic so as to reflect the impact of interest change by the time factor. Last but not least, we propose a method for predicting user interest.Theoretical analysis and experimental comparison prove the feasibility and efficiency of the proposed approach.