Diversely Regularized Matrix Factorization for Accurate and Aggregately Diversified Recommendation

When recommending personalized top-k items to users, how can we recommend them diversely while satisfying users’ needs? Aggregately diversified recommender systems aim a variety of across whole users without sacrificing the recommendation accuracy. They increase exposure opportunities various items, which in turn potential revenue sellers as well user satisfaction. However, it is challenging tackle aggregate-level diversity with matrix factorization (MF), one most common models, since skewed real-world data lead results MF. In this work, propose DivMF (Diversely Regularized Matrix Factorization), novel method for aggregately recommendation. exploits coverage regularizer and skewness consider an MF model diversify results. We also carefully designed training algorithm effective training. Extensive experiments on datasets show that gives state-of-the-art performance, improving up 34.7% similar level accuracy, 27.6% accuracy compared best competitors.