Attention on Global–Local Representation Spaces in Recommender Systems

In this study, we present a novel clustering-based collaborative filtering (CF) method for recommender systems. Clustering-based CF methods can effectively deal with data sparsity and scalability problems. However, most of them are applied to single representation space, which might not characterize complex user–item interactions well. We argue that the should be observed from multiple views characterized in an adaptive way. To address issue, leveraged global local properties construct spaces by learning various training datasets loss functions. An attention network was built generate blended according relative importance each pair, providing flexible way diverse interactions. Substantial experiments were evaluated on four popular benchmark datasets. The results show proposed is superior several where only one space considered.