Accuracy-Preserving Stability Regularization for Large-Scale Retail Demand Forecasting

arXiv cs.LG Papers

Summary

This paper introduces a training-time stability regularization penalty to improve forecast stability without sacrificing accuracy, evaluated on M5 retail demand data, showing improvements in Forecast Stability Score while maintaining RMSE within 0.72%.

arXiv:2607.13331v1 Announce Type: new Abstract: Retail demand forecasts are reused across replenishment, capacity, labor, and transportation planning cycles. Point-error objectives do not constrain abrupt movement between adjacent forecasts, while post-hoc smoothing acts only after model fitting. We ask whether a training-time penalty on consecutive within-series movement can improve horizontal forecast-path stability without materially changing point accuracy. The penalty is evaluated in a temporal-structured pipeline combining recent-demand embeddings with calendar, price, hierarchy, item, and store features. On selected M5 demand series at 1000, 3000, and 4000-series scales, the stability-aware hybrid model improves Forecast Stability Score over XGBoost by 6.91%, 6.66%, and 7.68%, respectively, while RMSE changes remain within 0.72% across three random seeds. Post-hoc exponential smoothing attains lower raw movement but incurs a larger RMSE cost; training-time regularization preserves more point accuracy and performs favorably under normalized stability. These findings extend forecast evaluation from point-error minimization toward an accuracy-stability trade-off perspective for operational retail forecasting.
Original Article
View Cached Full Text

Cached at: 07/16/26, 04:21 AM

# Accuracy-Preserving Stability Regularization for Large-Scale Retail Demand Forecasting
Source: [https://arxiv.org/abs/2607.13331](https://arxiv.org/abs/2607.13331)
[View PDF](https://arxiv.org/pdf/2607.13331)

> Abstract:Retail demand forecasts are reused across replenishment, capacity, labor, and transportation planning cycles\. Point\-error objectives do not constrain abrupt movement between adjacent forecasts, while post\-hoc smoothing acts only after model fitting\. We ask whether a training\-time penalty on consecutive within\-series movement can improve horizontal forecast\-path stability without materially changing point accuracy\. The penalty is evaluated in a temporal\-structured pipeline combining recent\-demand embeddings with calendar, price, hierarchy, item, and store features\. On selected M5 demand series at 1000, 3000, and 4000\-series scales, the stability\-aware hybrid model improves Forecast Stability Score over XGBoost by 6\.91%, 6\.66%, and 7\.68%, respectively, while RMSE changes remain within 0\.72% across three random seeds\. Post\-hoc exponential smoothing attains lower raw movement but incurs a larger RMSE cost; training\-time regularization preserves more point accuracy and performs favorably under normalized stability\. These findings extend forecast evaluation from point\-error minimization toward an accuracy\-stability trade\-off perspective for operational retail forecasting\.

## Submission history

From: Jize Li \[[view email](https://arxiv.org/show-email/3ddf02f8/2607.13331)\] **\[v1\]**Tue, 14 Jul 2026 23:31:00 UTC \(905 KB\)

Similar Articles

Optimizing ARDL Models for Retail Sales Forecasting and Fair Pricing

arXiv cs.LG

This paper proposes a fairness-aware pricing framework for retail food products using Autoregressive Distributed Lag (ARDL) models for sales forecasting and optimizes prices with Linear Programming and Simulated Annealing under CPI-based bounds to prevent consumer exploitation.