Research Of Personalized Recommendation Based On Weibo

As the mobile internet thriving,the development of social network is more rapid.In china,social products,eg Wechat,Sina Weibo,begin to draw more and more attention from people Weibo became an information platform because of two rules on it:First,restrict the number of words of Weibo,which makes every Weibo to be spread easily.Second,establish asymmetry relationship,which means that it is allowed that user follow another user single-direction and two users follow each otherAs Weibo becomes more and more prevailing,it confronts with the challenge of information overload.As the number of user increases,the relationship on Weibo become more complicated and the information produced become much huger.Hence,it’s a valuable research point to help user to find potential friend and useful information.Since the unique function design of Weibo,users on Weibo tend to adapt their social relationship to filter the content they receive Thus we mainly consider the problem of Weibo’s friend recommendation.This paper researches personalized recommendation based on Weibo.We emphasis the importance of recommendation scenarios,and propose to divide into 0-1 implicit and ordinary implicit scenario according to user’s life cycle on Weibo.First,we propose Constrained Pairwise Rank-biased Model to alleviate the problem of sparseness and ambiguity which exist in 0-1 implicit scenario.Second,on the context of ordinary implicit scenario,we propose a user behavior feature based recommendation algorithm.This algorithm provides hybrid recommendation result according the user demand which extracted from in user’s behavior feature.Finally,we design and implement a Weibo recommendation system,which performs the fast learning of recommendation model and provides a more persuasive from to display recommendation result.