About this Research Topic
With the rapid development of e-commerce and social media, recommendation systems have become indispensable tools for many online platforms. A personalized recommendation system is an advanced business intelligence platform based on massive data mining to help e-commerce websites provide fully personalized decision support and information services for their customers' shopping. In order to improve the efficiency of using information, solve the problem of information overload and optimize user experience, the recommendation system gives users personalized recommendations of potentially interesting information, products, etc. based on their needs and interest preferences, thus helping them find the information or products they need in a shorter period of time.
Since most of the data in recommendation systems can be represented by graph structures inherently (for example, user-item interaction data can be represented by bipartite graphs), user behavior sequences can be represented by directed graphs, and social relationships between users and knowledge graphs of items can also be used as structured side information. Therefore, to better utilize graph-structured data and improve model performance, many graph learning-based recommendation works have been proposed.
Early works utilize graph embedding techniques to model the relations between entities. Graph embedding techniques attempt to encode both users and items as continuous vectors in a shared space to better capture user preferences for items. This part of the work can be further divided into factorization-based methods, distributed representation-based methods, and neural embedding-based methods. Compared to earlier factorization-based methods, neural embedding-based methods can effectively capture non-linear relationships and easily incorporate rich side information. Inspired by the superior learning capability of GNNs on graph-structured data, large quantities of GNN-based recommendation models have emerged. By propagating information on the graph, higher-order structured information is encoded into the corresponding embeddings, instead of only acquiring first-order neighbors' information. Hence, by proposing this themed article collection, we hope to attract more interest from academia and industry to discuss all representative approaches, open problems, and future directions that can help take graph learning for recommender systems to new heights.
In this Research Topic, we are generally interested in applications, methodology, and application research focusing on graph learning for recommendation systems. Specifically, the following topics are preferable:
- Background of graph learning-based recommender systems.
- Challenges of applying graph learning techniques to recommendation systems.
- Categorizing existing works from different perspectives, such as stages, scenarios, goals, and applications.
- Introducing the representative models and the issues they address.
- Discussing insightful and promising future research directions in this area.
Since most of the data in recommendation systems can be represented by graph structures inherently (for example, user-item interaction data can be represented by bipartite graphs), user behavior sequences can be represented by directed graphs, and social relationships between users and knowledge graphs of items can also be used as structured side information. Therefore, to better utilize graph-structured data and improve model performance, many graph learning-based recommendation works have been proposed.
Early works utilize graph embedding techniques to model the relations between entities. Graph embedding techniques attempt to encode both users and items as continuous vectors in a shared space to better capture user preferences for items. This part of the work can be further divided into factorization-based methods, distributed representation-based methods, and neural embedding-based methods. Compared to earlier factorization-based methods, neural embedding-based methods can effectively capture non-linear relationships and easily incorporate rich side information. Inspired by the superior learning capability of GNNs on graph-structured data, large quantities of GNN-based recommendation models have emerged. By propagating information on the graph, higher-order structured information is encoded into the corresponding embeddings, instead of only acquiring first-order neighbors' information. Hence, by proposing this themed article collection, we hope to attract more interest from academia and industry to discuss all representative approaches, open problems, and future directions that can help take graph learning for recommender systems to new heights.
In this Research Topic, we are generally interested in applications, methodology, and application research focusing on graph learning for recommendation systems. Specifically, the following topics are preferable:
- Background of graph learning-based recommender systems.
- Challenges of applying graph learning techniques to recommendation systems.
- Categorizing existing works from different perspectives, such as stages, scenarios, goals, and applications.
- Introducing the representative models and the issues they address.
- Discussing insightful and promising future research directions in this area.
Keywords: Graph Neural Networks, Recommendation Systems, Knowledge Graph, Deep Learning, Representation Learning
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
