Collaborative Filtering Using Orthogonal Nonnegative Matrix Tri-factorization

Collaborative filtering aims at predicting a test user´s ratings for new items by integrating other like-minded users´ rating information. Traditional collaborative filter- ing methods usually suffer from two fundamental problems: sparsity and scalability. In this paper, we propose a novel framework for collaborative filtering by applying Orthogo- nal Nonnegative Matrix Tri-Factorization (ONMTF), which (1) alleviates the sparsity problem via matrix factorization; (2)solves the scalability problem by simultaneously cluster- ing rows and columns of the user-item matrix. Experimental results on benchmark data sets are presented to show that our algorithm is indeed more tolerant against both spar- sity and scalability, and achieves good performance in the meanwhile.