Research On The Modeling And Related Algorithms Of Collaborative Filtering Recommendation Based On Social Tagging

In the Internet era,social labels are begun to be used as one of the most important ways to organize,manage and share resources which classified and indexed resources.Social labels not only can express the characteristics of the resources themselves,but also reflect the interest preference of users.In order to improve the efficiency of personalized recommendation effectively,scholars have begun to combine them together.Collaborative filtering recommendation algorithm mainly depends on users' rating data of resources.But usually the user's score matrix is sparse,and bits of rating data cannot reflect the characteristics of users and resources very well.To solve the issue,this paper proposes an improved collaborative filtering recommendation algorithm based on the socialized label system.Therefore,this paper combines the label system with the collaborative recommendation system effectively.The algorithm based on the social tagging,embarks from the label,by splitting the three-dimensional relationships diagram of user-label-resources to two two-dimensional diagrams of users-tags and labels-resources.On the public data set Movie Lens,which compare the collaborative filtering recommendation algorithm based on the content of the traditional method,and solve the problem of traditional single interest model.Meanwhile,it also reduces the size of the rating matrix,and improve accuracy of recommendation.The writer complete the following work:(1)Propose a collaborative filtering Resource Recommendation Model Based on label system.Based on the inherent structure of the tag system,the tag sets is well integrated with the user sets and the resource sets,and the effective combination of the label system and the collaborative filtering recommendation algorithm based on the user and project is realized.(2)To analyze and discuss which based on the content of the the traditional collaborative filtering algorithm,introduce the key technology and model of collaborative filtering recommendation algorithm construction method,and in-depth inquiry of the problems existed in the sparse rating matrix,single interest model,and the score matrix is part of the algorithm is optimized.(3)The MovieLens public data sets is introduced and the experiment is carried out.Based on the MAE and HR/ARHR index,the two traditional algorithms and the two improved algorithms are evaluated.