Regularized Matrix Factorization with Cognition Degree for Collaborative Filtering

Abstract

Collaborative filtering is widely used technique in Recommender systems (RS) that are designed to deal with information overload problem. In particular, recently proposed methods based on Regularized Matrix Factorization (RMF) have shown promising results. However, these approaches focus on the user-item rating matrix, but ignore the significant influence of users’ preferences on items. In this paper, borrowed the idea of cognition degree, we propose a novel cognition degree-based RMF collaborative filtering model named CogRMF that model the interactions between users and items with users’ cognition degrees. In addition, Experiments on the real dataset Movielens 1M are implemented. Empirical outcomes show that the proposed model obtains significantly better results than other benchmark methods, such as user-based collaborative filtering (UCF), item-based collaborative filtering (ICF), cognition degree-based collaborative filtering (CDCF) and Regularized Matrix Factorization (RMF).