Explainable sequential recommendation using knowledge graphs

Knowledge Graphs have proven to be extremely valuable to recommender systems in recent years. By exploring the links within a knowledge graph, the connectivity between users and items can be discovered as paths, which provide rich and complementary information to user-item interactions. Leveraging this wealth of heterogeneous information for sequential recommendation is a challenging task, as it requires the ability to effectively encoding a diversity of semantic relations and connectivity patterns. To address the limitations of existing embedding-based and path-based methods for KG-aware recommendation, our work proposes a novel hybrid framework that naturally incorporates path representations with attentive weights derived from the knowledge graphs and sequential preference which links items with existing knowledge base into recommender systems to effectively recommend next item to a user. Our proposed model further employs a deep neural network to predict the interaction probabilities of a user and unseen items. Extensive experiments on real-world datasets illustrate that our approaches can give large performance improvements in a variety of scenarios, including movie, music and book recommendation.