Font Size: a A A

Research On One-class Collaborative Filtering Matrix Factorization Algorithm Fused With Hidden Information

Posted on:2022-07-18Degree:MasterType:Thesis
Country:ChinaCandidate:C T ZhangFull Text:PDF
GTID:2518306575466584Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Recommendation system is an effective means to solve the problem of information overload in the era of big data,and collaborative filtering is the key to the research of recommendation system.The collaborative filtering based on implicit feedback is called one-class collaborative filtering(OCCF),which is the focus and hotspot of current research.At the present stage,one-class collaborative filtering mainly relies on implicit feedback of users for matrix factorization,ignoring hidden information such as user information,item information,and rating information.Effective use of these hidden information helps improve the accuracy of recommendations.Therefore,considering the hidden information outside the 0-1 matrix and the differences of different types of hidden information,two matrix factorization algorithms are proposed to improve the efficiency of recommendation in this thesis.The main research contents are as follows:(1)Aiming at the problem that the matrix factorization in OCCF does not consider the contextual information of users and items,in this thesis,a matrix factorization algorithm integrating knowledge graph and representation learning is proposed.First,the algorithm uses knowledge graph and representation learning to represent the semantic information of users and items as low-dimensional vectors,and the implicit feedback matrix is improved by calculating the semantic distance between users and items.At the same time,the matrix factorization model is established based on the feature that entities in the knowledge graph interact with each other.Through the comparative experiment of different recommended list lengths,it is verified that the method in this thesis can optimize the characteristics of users and items from a semantic perspective.(2)Aiming at the problem that the matrix factorization in OCCF fails to effectively utilize the hidden information in the explicit feedback,in this thesis,we propose a matrix factorization algorithm integrating explicit feedback.The implicit feedback matrix is improved by using the information of rating and feature in the explicit feedback,and the implicit feedback matrix factorization model is established based on the feature that explicit feedback is more accurate to express the potential features.The algorithm is used to improve the s ALS model and w ALS model,and the comparative experiment of different recommended list lengths shows that the integration of explicit feedback can effectively optimize the interest model of user and improve the accuracy of recommendation.In summary,the two proposed algorithms in this thesis can enhance the accuracy of the recommendation system and verify the value of hidden information for recommendation system.At the same time,the experimental results show that the improved algorithms tend to be stable and general.
Keywords/Search Tags:recommendation system, collaborative filtering, one-class collaborative filtering, matrix factorization
PDF Full Text Request
Related items