Reproduce “F-FOMAML: GNN-Enhanced Meta-Learning for Peak Period Demand Forecasting with Proxy Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zexing Xu, Linjun Zhang, Sitan Yang, Rasoul Etesami, Hanghang Tong, Huan Zhang, Jiawei Han
Researchers from the University of Illinois Urbana-Champaign developed F-FOMAML, a meta-learning framework enhanced by Graph Neural Networks and Feature-wise Linear Modulation, to accurately predict demand for new products or during peak sales events with limited historical data. The model demonstrated improved forecasting accuracy on large-scale industrial datasets, including JD.com and a vending machine dataset.
Key insights
- Strategically leveraging "proxy data" (reflective sales patterns from similar entities during non-peak periods) can effectively mitigate data scarcity for peak event forecasting.
- GNN-enhanced representations, capturing cross-product relationships, provide valuable metadata that significantly improves the adaptability and accuracy of meta-learning models in data-limited scenarios.
- Theoretical analysis demonstrates that incorporating proxy data reduces excess risk and improves generalization, offering a better bias-variance trade-off in few-shot learning settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.