Enhanced Pairwise Learning for Personalized Ranking from Implicit Feedback

Yunzhou Zhang,Bo Yuan,Ke Tang

DOI: https://doi.org/10.1007/978-981-10-7179-9_46

2017-01-01

Abstract:One-class collaborative filtering with implicit feedback has attracted much attention, mainly due to the widespread of implicit data in real world. Pairwise methods have been shown to be the state-of-the-art methods for one-class collaborative filtering, but the assumption that users prefer observed items to unobserved items may not always hold. Besides, existing pairwise methods may not perform well in terms of Top-N recommendation. In this paper, we propose a new approach called EBPR, which relaxes the former simple pairwise preference assumption by further exploiting the hidden connection in observed items and unobserved items. EBPR can also be used as a basic method and has the extensive applicability, i.e., when combining our model with former pairwise methods, better performance can also be achieved. Empirical studies show that our algorithm outperforms the state-of-the-art methods on four real-world datasets.