Learning to Recommend with Hidden Factor Models and Social Trust Ensemble

Abstract

As one of the most successful approaches to building recommender systems, Collaborative Filtering (CF) uses the known preferences of a group of users to make recomm- endations or predictions of the unknown preferences for other users. The most successful approaches to CF are latent factor models, Latent Dirichlet Allocation (LDA) models, which directly profile both users and products, and trust-based collaborative filtering models, which analyze the connections among users. This paper introduces some innovations to both approaches. The factor, topic and trust models can now be smoothly merged, to build a more accurate combined model. Further accuracy improvements are achieved by extending the models to exploit both explicit and implicit feedback by the users. The methods are tested on the Yelp data, and the results are better than those previously published on that dataset.