Graph Neural Network for Product Recommendation on the Amazon Co-purchase Graph

TL;DR

This paper evaluates four Graph Neural Network architectures on the Amazon co-purchase graph, analyzing their performance, scalability, and practicality for real-world product recommendation systems.

Contribution

It provides a comparative analysis of LightGCN, GraphSAGE, GAT, and PinSAGE in the context of product recommendation, highlighting their strengths and trade-offs.

Findings

LightGCN showed superior scalability and performance.

GAT provided better semantic understanding but was more complex.

PinSAGE balanced performance and scalability effectively.

Abstract

Identifying relevant information among massive volumes of data is a challenge for modern recommendation systems. Graph Neural Networks (GNNs) have demonstrated significant potential by utilizing structural and semantic relationships through graph-based learning. This study assessed the abilities of four GNN architectures, LightGCN, GraphSAGE, GAT, and PinSAGE, on the Amazon Product Co-purchase Network under link prediction settings. We examined practical trade-offs between architectures, model performance, scalability, training complexity and generalization. The outcomes demonstrated each model's performance characteristics for deploying GNN in real-world recommendation scenarios.