Font Size: a A A

Collaborative Recommendation By Social Context Regularization And Item Context Regularization

Posted on:2013-01-20Degree:MasterType:Thesis
Country:ChinaCandidate:R Y CaiFull Text:PDF
GTID:2218330371458929Subject:Computer applications
Abstract/Summary:PDF Full Text Request
With the development of the web 2.0 technology, internet is playing a more and more important role in daily life. While e-commerce is widely spreading among people, the issue as to promote users with useful information from the expending information has become a big concern. And thus recommendation system was brought out in recent years.Traditional collaborative recommendation system analyzes user behavior and interest and then calculates the similarities among users, considers the ratings from similar users and finally and finally makes the recommendation. However, the sparse nature of actual user data results in the inaccuracy of similarity among users, and consequently the recommendation results is inaccurate. So, it has become a hot topic to overcome the issue brought by sparse nature of users in recommendation algorithm research area.The social net works such as twitter and face book are becoming more and more popular nowadays, they formed rich relationships among users. The relationships formed by interaction among users reflect the interest of user on one hand and also they represent the relationship among users. And thus it's a new approach to promote the recommendation result by getting useful information from these relationships.This paper introduces social context regularization and item context regularization into the factorization of low-rank matrix factorize, and proposes an approach for collaborative recommendation called collaborative recommendation by social context and item context regularization to improve the performance of collaborative recommendation. This method decomposes the user-rating matrix into a low-rank matrix A by the constraints on A with following conditions:1) A is a low-rank matrix; 2) Matrix A is satisfied with social context regularization, namely the uses within a same interest group tend to share the same preference, but the users from different groups may have different preferences; 3) Matrix A is satisfied with item context regularization, namely the items within a same interest group tend to share same features but the items from different groups may be absolutely different; Experimental results show that, CRSCR algorithm is superior to other collaborative recommendation algorithms.
Keywords/Search Tags:Social Context Regularization, Item Context Regularization, Low-rank Matrix Factorization, Latent Factor Model
PDF Full Text Request
Related items