Optimization-based approaches for maximizing aggregate recommendation diversity

Recommender systems are being used to help users find relevant items from a large set of alternatives in many online applications. Most existing recommendation techniques have focused on improving recommendation accuracy; however, diversity of recommendations has also been increasingly recognized in research literature as an important aspect of recommendation quality. This paper proposes several optimization-based approaches for improving aggregate diversity of top-N recommendations, including a greedy maximization heuristic, a graphtheoretic approach based on maximum flow or maximum bipartite matching computations, and an integer programming approach. The proposed approaches are evaluated using real-world movie rating data sets and demonstrate substantial improvements in both diversity and accuracy as compared to the recommendation reranking approaches, which have been introduced in prior literature for the purposes of diversity improvement and were used for baseline comparisons in our study. The paper also discusses the computational complexity and the scalability of the proposed approaches, as well as the potential directions for future work.