Researchers have introduced GNBAN, a Graph Neural Basis Attention Network designed for long-horizon forecasting across large sets of correlated entities, particularly in retail demand prediction. This novel architecture integrates heterogeneous graph representation learning with an interpretable decomposition head, allowing a single model to serve an entire catalog by representing retail data as a graph. GNBAN decomposes forecasts into trend, seasonal, and generic components, utilizing a per-basis attention mechanism to capture distinct temporal patterns while maintaining interpretability. Evaluations on the M5 Walmart and Favorita Grocery Sales benchmarks show GNBAN improves forecasting accuracy by 4-5% compared to existing graph-based methods. AI
IMPACT This model could improve the accuracy and interpretability of demand forecasting in large-scale retail operations.
RANK_REASON The cluster contains a research paper detailing a new model architecture.
- Favorita Grocery Sales
- GNBAN
- Graph Neural Basis Attention Network
- M5 Walmart
- Walmart
- alphaXiv
- CatalyzeX Code Finder for Papers
- Connected Papers
- DagsHub
- Gotit.pub
- Hugging Face
- Litmaps
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
