Pooling and Boosting for Demand Prediction in Retail: A Transfer Learning Approach
Bibliographic record
Abstract
Problem definition: How should retailers leverage aggregate (category) sales information for individual product demand prediction? Motivated by inventory risk pooling, we develop a new prediction framework that integrates category-product sales information to exploit the benefit of pooling. Methodology/results: We propose to combine data from different aggregation levels in a transfer learning framework. Our approach treats the top-level sales information as a regularization for fitting the bottom-level prediction model. We characterize the error performance of our model in linear cases and demonstrate the benefit of pooling. Moreover, our approach exploits a natural connection to regularized gradient boosting trees that enable a scalable implementation for large-scale applications. Based on an internal study with JD.com on more than 6,000 weekly observations between 2020 and 2021, we evaluate the out-of-sample forecasting performance of our approach against state-of-the-art benchmarks. The result shows that our approach delivers superior forecasting performance consistently with more than 9% improvement over the benchmark method of JD.com . We further validate its generalizability on a Walmart retail data set and through alternative pooling and prediction methods. Managerial implications: Using aggregate sales information directly may not help with product demand prediction. Our result highlights the value of transfer learning to demand prediction in retail with both theoretical and empirical support. Based on a conservative estimate of JD.com , the improved forecasts can reduce the operating cost by 0.01–0.29 renminbi (RMB) per sold unit on the retail platform, which implies significant cost savings for the low-margin e-retail business. History: This paper has been accepted as part of the 2023 Manufacturing & Service Operations Management Practice-Based Research Competition. Funding: This work was supported by the National Natural Science Foundation of China [Grant 71991462]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2022.0453 .
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".