SPADE-S: A Sparsity-Robust Foundational Forecaster
Malcolm Wolff, Matthew Li, Ravi Kiran Selvam, Hanjing Zhu, Kin G. Olivares, Ruijun Ma, Abhinav Katoch, Shankar Ramasubramanian, Mengfei Cao, Roberto Bandarra, Rahul Gopalsamy, Stefania La Vattiata, Sitan Yang, Michael M. Mahoney · 2025
Read SPADE-S: A Sparsity-Robust Foundational Forecaster by Malcolm Wolff, Matthew Li, Ravi Kiran Selvam, Hanjing Zhu, Kin G. Olivares, Ruijun Ma, Abhinav Katoch, Shankar Ramasubramanian, Mengfei Cao, Roberto Bandarra, Rahul Gopalsamy, Stefania La Vattiata, Sitan Yang, Michael M. Mahoney (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-21155.txt.
Description
Forecasting demand for millions of products is hard because most items sell rarely and in small amounts, while a few sell heavily. Malcolm Wolff and co-authors present SPADE-S in this 2025 paper, a forecasting system built to treat quiet and sparse series fairly instead of favoring the big sellers.
They trace rival models' bias to their scoring rules, sampling habits, and encodings, then design around each flaw. Across three retail data sets, from three million to seven hundred million series, their system beat top rivals, with the largest gains on sparse items. Better forecasts for the long tail mean fewer stock-outs and less waste.
Catalog record
| CREATOR | Malcolm Wolff, Matthew Li, Ravi Kiran Selvam, Hanjing Zhu, Kin G. Olivares, Ruijun Ma, Abhinav Katoch, Shankar Ramasubramanian, Mengfei Cao, Roberto Bandarra, Rahul Gopalsamy, Stefania La Vattiata, Sitan Yang, Michael M. Mahoney |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-21155