Research On Micro-blog Recommendation System Based On Trust Relationship

With the advent of the era of big data,exploring effective information for users has been becoming an important point of research in various fields.Micro-blog is one of the major social transaction platforms,having large data due to rapid spread rate and strong real-time.Therefore,recommending effective micro-blog information for users has become the focus of research in internal and external since the advent of Twitter.According to some problems of traditional recommendation system,such as the effect of the system based on content not good enough,the weak-interpretative ability of the system based on user's behavior,the limitation of the system based on user's friends.This paper proposes the Micro-blog recommendation system based on the trust relationship,the main purpose is that recommends micro-blog information to target user through the propagation of user's trust relationship,which published by target user's friends of friends,reliable and consistent with the recently emotion polarity of target user.The main contents of the this paper's study are as follows:This paper proposes the direct trust measurement model based on the follow relationship of user,the friends of user and the interaction of user through the analysis the follow condition and interaction of the user of micro-blog,and the simulated annealing algorithm is used to train the parameters of the model.This paper proposes the model for predicting indirect trust relation based on direct trust relation through analyze the propagation of user's trust.And building the user trust network based on trust relationship.This paper proposes a method of sentiment word extraction based on affective lexicon and SO-PMI algorithm,and realize the dynamic extension of the basic emotional word dictionary when analyzing the sentiment of micro-blog text.The influence of degree adverb and negative word is considerated when judging the sentiment polarity of micro-blog.Referencing the research results of user trust relationship and micro-blog sentiment analysis,This paper designed a micro-blog recommendation system based on trust relationship and sentiment analysis.In the light of the limitations of the system,this paper puts forward the reliable-rate of users,and uses the idea of Page Rank to calculate the user's reliable-rate.And then,combining the trust relation and reliable-rate of users to recommend the proper content to user.Finally,the parameters of the direct trust measurement model are determined by experiments,the experimental results of various data sets compared through the three indexes of precision,recall and F1-measure value.The results indicate that the effect of the proposed micro-blog recommendation system is better than the traditional micro-blog recommendation system based on content,user friendly or user behavior.