Research On Personalized Recommendation Problem Based On MicroBlog

With the rapid development of the Internet and Computer technology, theinformation age is facing a serious problem of information overload, which requiresfiltering irrelevant information through technical means, only to provide users withinformation about their interest in order to achieve personalized recommendationinformation. How to effectively mining the user’s individual interest is key topersonalized recommendations.In recent years, with the microblog rapid development and wide application,microblog has become an important social media. Microblog is similar interests as alink network of social platforms, user interests, behavior, etc. can be expressed bymicroblog. Microblog rely on its content with simplicity, terminal scalability,openness and platform features such as low-threshold quickly won the favor of users,has developed into a personal expression and social interaction, event participationand an important platform for content sharing and profound impact on the economicand social development. Therefore, further study of microblog, tap the user’spersonalized interests, will provide a personalized recommendation of potentialpathways and channels.Microblog is significantly differ from the other social network, which hasfeatures such as social media nature, scale, noise data diverse, non-linear, the rapidspread of evolution, as well as multi-relational, and must be used in different waysfrom other social networks to analyze mining. In this paper, the goal is mining userinterests, deeply studies on the timeliness of content microblog, microblog socialcharacteristics of interest and update of user interest and other issues, the maininnovation of this paper and contributions are as follows:(1) For the timeliness issue of microblog content, this paper presented T-LDAusers based on their interests microblog mining algorithms to effectively solve theuser’s interest migrate over time occurrence, may no longer be concerned about theinterest of the original problem. This paper presents a time-sensitive T-LDA algorithm itself mainly mining userinterest from three aspects, one is interested in long-term users, which is the subjectof information for a long time user microblog content repeatedly involved; the otheris the user recent interest in the theme of the information that the user in the nearfuture before its microblog content is not involved; and the third is the user’s expiredinterest. Considering the interest in the theme of time and weight, and tap the user’sown interest. Experimental results show that the algorithm can dig out more satisfiedusers interested in the topic itself.(2) For microblog social characteristics, we propose a social based on userinterest mining algorithms FInterest, it is an effective solution to the user of socialinterest are not of concern.This paper presents a user followee interest FInterest mining algorithm,firstlyanalyze the followee of the user, mining the special followee list of the user,secondly mining the long-term interests of these special followee, it represents along-term and stable focus on followee interest category; finally dig all usersimmediate interest of all of the followee, on behalf of the user may be interested inareas of interest in migration. To achieve an effective social interest concerns.Experimental results show that, FInterest algorithm for mining user followeeinterested can obtain more effective user interest topics.(3) For can not timely access to the latest user interest problem, a feedbackmechanism to evaluate the introduction of user interest model update algorithm cannot solve the problem in time to update the user feedback.This paper presents an evaluation of feedback mechanism to introduce userinterest model update algorithm, which allows users of the system by therecommended information or links to read the feedback rating feedback, record eachlink corresponds to the theme, and then the corresponding user topics recommendedvalues gain or reduction, the effective realization of user feedback to update.Experimental results show that the feedback mechanism to evaluate user interestmodel mining not only can better track users interested in migrating the latest trends,get more real user interest, but also more accurate information to achieve a personalized recommendation.In this paper, deeply analysis of the characteristics of microblog, timeawareness for its content, the social characteristics of the micro-blog of interest andupdate user interest and other issues related to research, by combining the use of anopen platform Sina Weibo and Douban open platform design the correlationalgorithm validation and testing, to prepare a solid theoretical foundation to carry outpersonalized recommendations based on the study of microblog.