Personalized recommendation with implicit feedback via learning pairwise preferences over item-sets

Preference learning is a fundamental problem in various smart computing applications such as personalized recommendation. Collaborative filtering as a major learning technique aims to make use of users’ feedback, for which some recent works have switched from exploiting explicit feedback to implicit feedback. One fundamental challenge of leveraging implicit feedback is the lack of negative feedback, because there is only some observed relatively “positive” feedback available, making it difficult to learn a prediction model. In this paper, we propose a new and relaxed assumption of pairwise preferences over item-sets, which defines a user’s preference on a set of items (item-set) instead of on a single item only. The relaxed assumption can give us more accurate pairwise preference relationships. With this assumption, we further develop a general algorithm called CoFiSet (collaborative filtering via learning pairwise preferences over item-sets), which contains four variants, CoFiSet(SS), CoFiSet(MOO), CoFiSet(MOS) and CoFiSet(MSO), representing “Set vs. Set,” “Many ‘One vs. One’,” “Many ‘One vs. Set”’ and “Many ‘Set vs. One”’ pairwise comparisons, respectively. Experimental results show that our CoFiSet(MSO) performs better than several state-of-the-art methods on five ranking-oriented evaluation metrics on three real-world data sets.