Social Recommendation With Local Low Rank Matrix Approximation.

Huan Zhao,Quanming Yao,Dik Lun Lee
2017-01-01
Abstract:In the recommendation literature, social connections have been successfully incorporated into the traditional recommendation methods, especially the most popular state-of-art one Matrix Factorization (MF). MF is based on the assumption that the user-item rating matrix is low rank and thus can be decomposed into two smaller matrices representing the user and item latent features. Recently, instead of assuming that the rating matrix is low rank, Local LOw Rank Matrix Approximation (LLORMA) has been proposed based on a novel assumption that the rating matrix is com- posed of a set of low-rank submatrices, termed local low rank. Instead of factorizing the original matrix, these low rank submatrices are factorized independently to approximate the original rating matrix. Experimental results have shown that LLORMA improves the recommending performance significantly comparing to standard MF. Unfortunately, LLORMA does not exploit any social information. In this paper, we propose a novel model called Social Local Weighted Matrix Factorization (SLWMF ), which is the first work to model social recommendation in the local low rank framework. SLWMF is based on the intuitive idea that the recommendation performance can surely be improved if we can intelligently integrate social connections into the local low rank framework.
What problem does this paper attempt to address?