Graph-based context-aware collaborative filtering

Abstract

Context-aware recommender systems (CARS) are specially designed to take into account the contextual conditions under which a user experiences an item, with the goal of generating improved recommendations. A known difficulty when constructing recommender systems is data sparseness, which reduces the effectiveness of collaborative filtering algorithms. While using contextual information provides fine-grained signals for the recommendation process, it makes the data even sparser and increases the computational complexity. In this paper, we present a method for making context-aware recommendations, which is less sensitive to data sparseness. The proposed method exploits the transitivity of the interactions between users and items on the user-item graph to augment the direct interactions, thus reducing the negative effect of sparse data. Based on graph transitivity we introduce a new graph-based association measure that we use as a measure of similarity between two users or two items in nearest neighbor recommendation methods. This combination of graph-based similarity measure with nearest neighbor methods allows considering more contextual conditions at a lower risk of being affected by data sparseness caused by additional contextual dimensions. We experimentally evaluated the proposed method on three contextually-tagged data sets. The results show that our method outperforms several baselines and state-of-the-art context-aware recommendation methods.