Interest Mining And Modeling For Micro-bloggers Of Micro-blog

As a light-weight blog, the micro-blog attracts a large amount of Internet users in a short period of time due to its convenience and ease of use. The contents and the information in micro-blogs can be spread virally and rapidly through forwarding by the micro-bloggers, which brings in great business opportunities. Thus the micro-blogging sites have already become an important advertising platform. Advertising is a key means to make profit for micro-blogging sites and its accuracy would affect the user experience, the advertising effect to a large extent. The user interest model of micro-bloggers is the basis of advertising accuracy and thus the study of the user interest modeling is meaningful for both the academic research and the development micro-blogging sites.To model the user interest of the micro-bloggers accurately, the thesis mainly focused on the following two problems. For one thing, to choose the appropriate information to denote the user interest based on the characteristics of the micro-blogging sites and the micro-bloggers. For the other thing, to build the user interest model. The modeling process includes obtaining the user interest, classification of the user interest, and the calculation of long-and short-term interest. We first analyzed the traits of the micro-blogs and then their users, and next figured out the accurate information to describe the user interest, and then addressed the user interest modeling method for micro-bloggers, and finally testified the method.Based on the data from Sina Weibo, the thesis carried out research mainly into the following two aspects:First, we analyzed the contents of the micro-blogs and the behaviors of the micro-bloggers to get the appropriate information to describe the user interest through study of the relationship between the interest and the behaviors. We found that a few selective micro-blogs and the existing user tags can hardly describe the user interest of micro-blogger accurately and decided to choose the special micro-bloggers, mainly the celebrity ones followed by the specific user as his or her source of user interest. Second, we extracted the traits of the micro-bloggers’ user interest and designed the user interest modeling method. By designing the evaluation indexes and applying the method to the designed cases, we testified its feasibility, the accuracy of the obtained user interest model, and the sensitivity in assessing the interest drift.