Social-aware Matrix Factorization for Recommender Systems
Daniel Karl I. Weidele · KOPS (University of Konstanz) · 2013
We review and categorize early approaches of collaborative filtering, before moving towards social-aware matrix factorization models for rating prediction, which we will theoretically compare to each other and to the state of the art model SVD++. We derive a generic social-aware factorization model and show how to improve runtime complexities of social-aware matrix factorization models in general. Moreover we discuss various trust metrics to exploit social network information and propose the application of PageRank as a new alternative in this context. Finally we provide a practical evaluation of presented approaches.