Authors: Haitong Luo, Xuying Meng, Suhang Wang, Hanyun Cao, Weiyao Zhang, Yequan Wang, Yujun Zhang

Published on: January 04, 2024

Impact Score: 8.07

Arxiv code: Arxiv:2401.0213

Summary

• What is new: The discovery that complementary relationships in items involve both low-frequency and mid-frequency spectral components, and the use of spectral-based GNNs to model these relationships.
• Why this is important: Existing models fail to effectively balance relevance and dissimilarity in complementary item relationships, impacting recommendation accuracy.
• What the research proposes: A novel Spectral-based Complementary Graph Neural Network (SComGNN) that uses spectral graph convolutions with low-pass and mid-pass filters, coupled with a two-stage attention mechanism.
• Results: SComGNN significantly outperforms baseline models on four e-commerce datasets, proving its effectiveness in modeling complementary relationships.

Technical Details

Technological frameworks used: Spectral-based Graph Neural Networks

Models used: Spectral graph convolutional networks with low-pass and mid-pass filters, two-stage attention mechanism

Data used: Four e-commerce datasets

Potential Impact

E-commerce platforms and online recommendation systems could greatly benefit, enhancing user experience and potentially increasing sales through better product suggestions.

Want to implement this idea in a business?

We have generated a startup concept here: Complementify.