Ripple Knowledge Graph Convolutional Networks For Recommendation Systems

TL;DR

RKGCN addresses information overload by jointly enriching user and item representations with knowledge graphs in an end-to-end framework. It introduces a two-branch architecture: User Preference Aggregation to capture dynamic user interests and Entity Enhancement to inject KG structural signals into item representations, with final scoring via an inner product. Across three real-world datasets, RKGCN outperforms five baselines in AUC and ACC, particularly when knowledge graphs are dense, demonstrating the value of simultaneous user- and item-side KG utilization. The work advances KG-powered recommender systems and points to future directions in scalable sampling, robustness to KG noise, and adversarial improvements.

Abstract

Using knowledge graphs to assist deep learning models in making recommendation decisions has recently been proven to effectively improve the model's interpretability and accuracy. This paper introduces an end-to-end deep learning model, named RKGCN, which dynamically analyses each user's preferences and makes a recommendation of suitable items. It combines knowledge graphs on both the item side and user side to enrich their representations to maximize the utilization of the abundant information in knowledge graphs. RKGCN is able to offer more personalized and relevant recommendations in three different scenarios. The experimental results show the superior effectiveness of our model over 5 baseline models on three real-world datasets including movies, books, and music.

Figures (9)

• Figure 1: An example of a recommendation system for CF.
• Figure 2: Examples of rating and purchase/click matrices for a movie recommendation system.
• Figure 3: An example of a knowledge graph-enhanced movie recommendation system.
• Figure 4: An example of user’s preference set.
• Figure 5: An example of an entity update. An entity embedding is updated from peripheral nodes to central nodes by a loop operation.