Graph-Based Recommendation

A gentle introduction to recommender systems working with and on graphs.

Abstract

Most of today’s data is structured as nodes connected through edges forming graphs, from molecules and drugs to traffic flows and social networks. The same applies to data stored on popular online platforms (e.g., Amazon, Netflix, X, Booking.com) which host vast and heterogenous catalogues of products or services enjoyed daily by different customer segments.

In this respect, recommender systems are complex algorithms designed to match users preferences and tastes to products and services, easing our navigation experience while improving the revenues of the companies behind those online platforms.

However, traditional recommendation approaches fail to perceive the user-item recommendation data as graph structures, disregarding a large amount of meaningful properties and characteristics that could be exploited to improve the recommendation quality.

Conversely, graph-based recommender systems leverage the user-item graph topology at multiple distance hops to derive finer-grained preference patters of users interacting with items.

The scope of this lecture is to provide the useful background notions regarding traditional recommendation, and the rationales behind the more improved and recent approaches from graph-based recommendation. Then, it presents pioneer solutions in graph-based recommendation leveraging, among others, random-walk techniques and graph neural networks.

Contents

• The recommendation task and traditional models
• Graph-based recommendation
• Background notions on graph neural networks
• Current directions in graph neural networks for recommendation

Previous iterations of the lecture

• Invited workshop at Cognism.com, invited by Felice Merra (April 2025)
• Guest lecture as part of the course of "Machine Learning in Network Science" held at CentraleSupélec and coordinated by Prof. Fragkiskos Malliaros (April 2025)

My works on graph-based recommendation

2025

• ContextGNN goes to Elliot: Towards Benchmarking Relational Deep Learning for Static Link Prediction (aka Personalized Item Recommendation). A Ariza-Casabona, N Kanakaris, D Malitesta. arXiv preprint arXiv:2503.16661
• Advances on Graph-Based Approaches in Information Retrieval. L Boratto, D Malitesta, M Marras, G Medda, C Musto, E Purificato. Communications in Computer and Information Science 2197, 1-98

2024

• A Novel Evaluation Perspective on GNNs-based Recommender Systems through the Topology of the User-Item Graph. D Malitesta, C Pomo, VW Anelli, ACM Mancino, T Di Noia, E Di Sciascio. The 18th ACM Conference on Recommender Systems (RecSys’24)
• Do We Really Need to Drop Items with Missing Modalities in Multimodal Recommendation?. D Malitesta, E Rossi, C Pomo, T Di Noia, FD Malliaros. The 33rd ACM International Conference on Information and Knowledge Management (CIKM'24)
• Uplift Modeling Under Limited Supervision. G Panagopoulos, D Malitesta, FD Malliaros, J Pang. Joint European Conference on Machine Learning and Knowledge Discovery in Databases (ECML-PKDD'24)
• Graph Neural Networks for Recommendation leveraging Multimodal Information. D Malitesta. ACM SIGIR Forum 58 (1)
• First International Workshop on Graph-Based Approaches in Information Retrieval (IRonGraphs 2024). L Boratto, D Malitesta, M Marras, G Medda, C Musto, E Purificato. The 46th European Conference on Information Retrieval (ECIR'24)

2023

• Graph Neural Networks for Recommendation: Reproducibility, Graph Topology, and Node Representation. D Malitesta, C Pomo, T Di Noia. The 2nd Learning on Graphs Conference (LoG'23)
• KGTORe: Tailored Recommendations through Knowledge-aware GNN Models. ACM Mancino, A Ferrara, S Bufi, D Malitesta, T Di Noia, E Di Sciascio. The 17th ACM Conference on Recommender Systems (RecSys'23)
• Challenging the Myth of Graph Collaborative Filtering: a Reasoned and Reproducibility-driven Analysis. VW Anelli, D Malitesta, C Pomo, A Bellogin, E Di Sciascio, T Di Noia. The 17th ACM Conference on Recommender Systems (RecSys'23)
• An Out-of-the-Box Application for Reproducible Graph Collaborative Filtering extending the Elliot Framework. D Malitesta, C Pomo, VW Anelli, T Di Noia, A Ferrara. The 31st ACM Conference on User Modeling, Adaptation and Personalization (UMAP'23)
• Auditing Consumer- and Producer-Fairness in Graph Collaborative Filtering. VW Anelli, Y Deldjoo, T Di Noia, D Malitesta, V Paparella, C Pomo. The 45th European Conference on Information Retrieval (ECIR'23)

2022

• Reshaping Graph Recommendation with Edge Graph Collaborative Filtering and Customer Reviews. VW Anelli, Y Deldjoo, T Di Noia, A Ferrara, D Malitesta, C Pomo. Workshop on Deep Learning for Search and Recommendation (CIKM'22)
• How Neighborhood Exploration influences Novelty and Diversity in Graph Collaborative Filtering. VW Anelli, Y Deldjoo, T Di Noia, A Ferrara, D Malitesta, C Pomo. 2nd Workshop on Multi-Objective Recommender Systems (RecSys'22)

Other details

Coming soon: lecture notes (stay tuned!)
Lecturer: Daniele Malitesta
Email: [email protected]