Microblog Recommendation Method Based On Information Extension And Trust Perception

With the rapid development of the Internet,the popularity and development rate of social networks have been greatly improved.As one of the mainstream social network platforms,microblog has successfully attracted a large number of users due to its features of simple operation and fast information update.With the increase of the number of users,the information of microblog grows explosively,which makes the information overload problem of microblog platform increasingly obvious.In order to avoid the interference of information overload,microblog recommendation has gradually become one of the current research hotspots.Microblog recommendation is similar to other recommendations.Its essence mainly includes user interest information,Microblog information to be recommended,and the calculation of the matching degree between the two.However,compared with recommendations in other fields,Microblog recommendations have their own significant characteristics,such as: users' interest is not obvious,data sparsity exists in microblog texts,and the importance of microblog itself influences the recommendation effect.According to the characteristics of microblog recommendation,this paper proposes a new microblog recommendation method based on VSM(Vector Space Model).This method mainly involves three aspects:(1)Expand user interests: Firstly,the user's interest information is obtained based on the user's history Microblog text.Secondly,based on the interaction between users,the CD(Community Division)method is used to obtain the community interest of the user's community.Finally,the community interest is used to expand the user's own interest,and then get the user's interest expression method;(2)Extend the original recommendation Microblog text: Firstly,the initial information of microblog was obtained based on the text content of the initial microblog.Then,based on the micro-blog search engine,the pseudo-correlation feedback method was used to obtain the extended corpus of microblog and obtain the effective extended information.Finally,the original microblog was extended with the extended information,and then the presentation method of the microblog to be recommended was obtained;(3)The user trust is integrated into the matching calculation of user interest and the initial information of microblog.Firstly,the user's trust on the recommended microblog is quantified based on the user's familiarity trust,similarity trust and microblog text influence.Then,the user's trust in the recommended microblog is integrated into the similarity calculation of the user's interest and the microblog to be recommended,so as to obtain the microblog recommendation method integrating trust perception.In this paper,12956 pieces of real sina microblog data were crawled through the network to verify the effectiveness and rationality of the research work in this paper.Firstly,microblog data were cleaned and preprocessed,and a data set for verifying microblog recommendation was constructed according to experimental requirements.Then,based on the constructed data set,the recommendation accuracy and ranking of different microblog recommendation methods were compared.The experimental results show that: compared with the phased work in this paper,the microblog recommendation method finally proposed in this paper has higher Precision(P@15)and mean Average Precision(MAP);Compared with basic research and mainstream research in the field of microblog recommendation,the microblog recommendation method proposed in this paper also has higher P@15 and MAP.