Relational Collaborative Filtering

Abstract

Existing item-based collaborative filtering (ICF) methods leverage only the relation of collaborative similarity -i.e., the item similarity evidenced by user interactions like ratings and purchases. Nevertheless, there exist multiple relations between items in real-world scenarios, e.g., two movies share the same director, two products complement with each other, etc. Distinct from the collaborative similarity that implies co-interact patterns from the user's perspective, these relations reveal fine-grained knowledge on items from different perspectives of meta-data, functionality, etc. However, how to incorporate multiple item relations is less explored in recommendation research.In this work, we propose Relational Collaborative Filtering (RCF) to exploit multiple item relations in recommender systems. We find that both the relation type (e.g., shared director) and the relation value (e.g., Steven Spielberg) are crucial in inferring user preference. To this end, we develop a two-level hierarchical attention mechanism to model user preference -the first-level attention discriminates which types of relations are more important, and the second-level attention considers the specific relation values to estimate the contribution of a historical item. To make the item embeddings be reflective of the relational structure between items, we further formulate a task to preserve the item relations, and jointly train it with user preference modeling. Empirical results on two real datasets demonstrate the strong performance of RCF 1 . Furthermore, we also conduct qualitative analyses to show the benefits of explanations brought by RCF's modeling of multiple item relations. CCS CONCEPTS• Information systems → Recommender systems; Retrieval models and ranking; Novelty in information retrieval.

Seismogenic structures of the 2006 ML4.0 Dangan Island earthquake offshore Hong Kong

“…We summarize the statistics of our evaluation datasets in Table I. Following the settings in [42], [56], we generate the item-wise relations with external knowledge (e.g., product categories, business genres) from the item side.…”

“…In real-world recommendation scenarios, there typically exist dependencies across items, e.g., product categories/functionality, spatial similarities of venues [17], [37]. Such rich semantic relatedness among items can help explore their latent dependencies, which is helpful to understand complex interests of users [42], [56]. As a result, user's preference over different items may not only be affected by his/her social connections, but also be inferred from the fine-grained relational knowledge on items.…”

Graph attentive matrix factorization for social recommendation

“…To reduce the time complexity of ranking all items for each user, we randomly sample 1000 unrated items for each user and rank them alongside the ground‐truth items. We use two common ranking‐based metrics, Hit Ratio (HR@N) and Normalized Discounted Cumulative Gain (NDCG@N; He et al, 2017; He et al, 2018; Xin et al, 2019), to measure the performance of the ranked lists. Intuitively, HR@N estimates the percentage of top‐ranked items that match the user's preferences, while NDCG@N accounts for the position of popular items by assigning higher scores to the top‐ranked items.…”

Multi-Behavior Graph Neural Networks for Recommender System

“…Some of studies follow this research line to enable the nonlinear feature interactions with the multi-Layer feed-forward network, such as NCF [11] and DMF [47]. To consider item relational data into the CF model, the relational collaborative filtering (RCF [46]) framework designs neural two-stage attention mechanism to enhance the item embedding process.…”