Beyond Similarity: Relation-Based Collaborative Filtering

Guannan Liu, Liang Zhang, Junjie Wu

Abstract: Given the effectiveness and ease of use, Item-based Collaborative Filtering (ICF) methods have been broadly used in industry and are widely investigated in recent years. The key of ICF lies in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">similarity</i> measurement between items, which however is a coarse-grained numerical value that can hardly capture users’ fine-grained preferences toward different attributed aspects of items. In this paper, we propose a model called REDA (Relation Embedding with Dual Attentions) to address this challenge, based on which a new paradigm called Relation-based Collaborative Filtering is designed for high-performance recommendation. REDA is essentially a deep neural network model that employs an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">item relation embedding</i> scheme for inter-item relations representation. It features in multi-decomposed item embedding with dual-attention refinement and employs a novel relation-wise optimization scheme for end-to-end learning. A <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">relational user embedding</i> is then proposed by aggregating item relation embeddings between all purchased items of a user, which not only profiles users’ fine-grained preferences but also alleviates the data sparsity problem. Extensive experiments are conducted on six real-world datasets and the proposed REDA is shown to outperform ten state-of-the-art methods. In particular, REDA shows great robustness against data and relation sparsity, the ability to learn explainable item aspects, and the potential for large-scale recommendation.