On the Impact of Graph Neural Networks in Recommender Systems: A Topological Perspective

TL;DR

This work reframes graph-based recommender systems through a topology-centered lens, arguing that both data graph structure and GNN architectural choices jointly shape performance. It develops a formal pipeline and a taxonomy of eleven models, then defines classical and topological dataset properties to explain when and why GNNs excel. Through an explanatory framework that uses graph sampling and linear regression, the authors link dataset topology (e.g., node degree, clustering, assortativity) to recommendation accuracy, revealing that node degree is the most explicit driver while higher-order topology explains model differences. The study provides practical guidance for topology-aware model selection and evaluation, and outlines theoretical, data-centric, and methodological challenges for advancing topology-aware recommender systems. Overall, the work highlights how topological insights can predict performance, guide architecture design, and inform fair and robust deployment of GNN-based recommendations.

Abstract

In recommender systems, user-item interactions can be modeled as a bipartite graph, where user and item nodes are connected by undirected edges. This graph-based view has motivated the rapid adoption of graph neural networks (GNNs), which often outperform collaborative filtering (CF) methods such as latent factor models, deep neural networks, and generative strategies. Yet, despite their empirical success, the reasons why GNNs offer systematic advantages over other CF approaches remain only partially understood. This monograph advances a topology-centered perspective on GNN-based recommendation. We argue that a comprehensive understanding of these models' performance should consider the structural properties of user-item graphs and their interaction with GNN architectural design. To support this view, we introduce a formal taxonomy that distills common modeling patterns across eleven representative GNN-based recommendation approaches and consolidates them into a unified conceptual pipeline. We further formalize thirteen classical and topological characteristics of recommendation datasets and reinterpret them through the lens of graph machine learning. Using these definitions, we analyze the considered GNN-based recommender architectures to assess how and to what extent they encode such properties. Building on this analysis, we derive an explanatory framework that links measurable dataset characteristics to model behavior and performance. Taken together, this monograph re-frames GNN-based recommendation through its topological underpinnings and outlines open theoretical, data-centric, and evaluation challenges for the next generation of topology-aware recommender systems.

Figures (3)

• Figure 1: Computed degree distributions for all the selected datasets in this monograph. On the $x$ and $y$ axes, we report the degree values (both user and item side) and the probability distribution (in log scale), respectively.
• Figure 2: Pearson correlation of the selected characteristics and datasets in the study by DBLP:conf/recsys/MalitestaPAMNS24. Many values in $[-0.5, 0.5]$ indicate loose correlation.
• Figure 3: Visual representation of the impact of dataset characteristics on the recommendation performance (Recall@20) of GNNs-based recommender systems, for each dataset/model setting in DBLP:conf/recsys/MalitestaPAMNS24. Bar plot length and direction represent the impact magnitude and whether there is a direct/inverse correspondence between characteristic and performance. Finally, the darker the bar plots, the higher their statistical significance.