Risk-Aware Decision-Focused Demand Forecasting for Multi-Channel Inventory Optimization under Promotion and Stock-Out Constraints
DOI:
https://doi.org/10.54097/ag16zy57Keywords:
Demand forecasting, decision-focused learning, inventory optimization, conditional value-at-risk, censored demand, omnichannel retailing, promotions.Abstract
Retail replenishment couples a forecasting problem with a constrained stochastic allocation problem, yet forecasters are still trained to minimise statistical error rather than the cost of the decisions they induce. The mismatch is aggravated by three features of real multi-channel retailing: channels of one region draw on a shared, capacity-limited stock; promotions inflate both demand and the penalty of a stock-out; and observed sales are censored precisely when a stock-out occurs. We propose RA-DFL, which trains a probabilistic forecaster through a differentiable multi-channel allocation layer that minimises a convex combination of expected cost and conditional value-at-risk under a shared capacity, a fill-rate requirement and omnichannel recourse. The layer solves the recourse stage in closed form and is optimised by a smoothed projected-gradient scheme with an exact line search, attaining a mean optimality gap of 0.32% against derivative-free solvers. A censoring-consistent decision loss removes the spurious overage signal created by truncated sales, and the forecaster emits a state-conditional risk level. On the real M5 (Walmart) and Stallion distribution panels we first establish a negative result: attaching a CVaR objective to strong predict-then-optimise forecasts raises realised cost by 9.7% and 11.2% respectively. Training through the risk-aware decision instead lowers mean cost by 3.7% and pool-level tail cost by 0.2% against the strongest predict-then-optimise pipeline on M5, and dominates its mean-risk frontier. Ablations show the censoring component helps significantly on the panel with 12% censoring and not on the panel with 1% censoring, evidence that it acts through the intended mechanism.
Downloads
References
[1] Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2022). M5 accuracy competition: Results, findings, and conclusions. International Journal of Forecasting, 38(4), 1346–1364. DOI: https://doi.org/10.1016/j.ijforecast.2021.11.013
[2] Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2022). The M5 competition: Background, organization, and implementation. International Journal of Forecasting, 38(4), 1325–1336. DOI: https://doi.org/10.1016/j.ijforecast.2021.07.007
[3] Makridakis, S., Spiliotis, E., Assimakopoulos, V., Chen, Z., Gaba, A., Tsetlin, I., & Winkler, R. L. (2022). The M5 uncertainty competition: Results, findings and conclusions. International Journal of Forecasting, 38(4), 1365–1385. DOI: https://doi.org/10.1016/j.ijforecast.2021.10.009
[4] Petropoulos, F., Wang, X., & Disney, S. M. (2019). The inventory performance of forecasting methods: Evidence from the M3 competition data. International Journal of Forecasting, 35(1), 251–265. DOI: https://doi.org/10.1016/j.ijforecast.2018.01.004
[5] Elmachtoub, A. N., & Grigas, P. (2022). Smart “predict, then optimize”. Management Science, 68(1), 9–26. DOI: https://doi.org/10.1287/mnsc.2020.3922
[6] Mandi, J., Kotary, J., Berden, S., Mulamba, M., Bucarey, V., Guns, T., & Fioretto, F. (2024). Decision focused learning: Foundations, state of the art, benchmark and future opportunities. Journal of Artificial Intelligence Research, 80, 1623–1701. DOI: https://doi.org/10.1613/jair.1.15320
[7] Donti, P., Amos, B., & Kolter, J. Z. (2017). Task based end to end model learning in stochastic optimization. In Advances in Neural Information Processing Systems 30.
[8] Amos, B., & Kolter, J. Z. (2017). OptNet: Differentiable optimization as a layer in neural networks. In Proceedings of the International Conference on Machine Learning (pp. 136–145).
[9] Agrawal, A., Amos, B., Barratt, S., Boyd, S., Diamond, S., & Kolter, J. Z. (2019). Differentiable convex optimization layers. In Advances in Neural Information Processing Systems 32.
[10] Wilder, B., Dilkina, B., & Tambe, M. (2019). Melding the data decisions pipeline: Decision focused learning for combinatorial optimization. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 33, pp. 1658–1665). DOI: https://doi.org/10.1609/aaai.v33i01.33011658
[11] Berthet, Q., Blondel, M., Teboul, O., Cuturi, M., Vert, J. P., & Bach, F. (2020). Learning with differentiable perturbed optimizers. In Advances in Neural Information Processing Systems 33 (pp. 9508–9519).
[12] Bertsimas, D., & Kallus, N. (2020). From predictive to prescriptive analytics. Management Science, 66(3), 1025–1044. DOI: https://doi.org/10.1287/mnsc.2018.3253
[13] El Balghiti, O., Elmachtoub, A. N., Grigas, P., & Tewari, A. (2019). Generalization bounds in the predict then optimize framework. In Advances in Neural Information Processing Systems 32.
[14] Hu, Y., Kallus, N., & Mao, X. (2022). Fast rates for contextual linear optimization. Management Science, 68(6), 4236–4245. DOI: https://doi.org/10.1287/mnsc.2022.4383
[15] Schutte, N., Postek, K., & Yorke Smith, N. (2024). Robust losses for decision focused learning. In Proceedings of the 33rd International Joint Conference on Artificial Intelligence (IJCAI) (pp. 4868–4875).
[16] Tang, B., & Khalil, E. B. (2024). PyEPO: A PyTorch based end to end predict then optimize library for linear and integer programming. Mathematical Programming Computation, 16, 297–335. DOI: https://doi.org/10.1007/s12532-024-00255-x
[17] Ban, G. Y., & Rudin, C. (2019). The big data newsvendor: Practical insights from machine learning. Operations Research, 67(1), 90–108. DOI: https://doi.org/10.1287/opre.2018.1757
[18] Oroojlooyjadid, A., Snyder, L. V., & Takáč, M. (2020). Applying deep learning to the newsvendor problem. IISE Transactions, 52(4), 444–463. DOI: https://doi.org/10.1080/24725854.2019.1632502
[19] Qi, M., Shi, Y., Qi, Y., Ma, C., Yuan, R., Wu, D., & Shen, Z. J. (2023). A practical end to end inventory management model with deep learning. Management Science, 69(2), 759–773. DOI: https://doi.org/10.1287/mnsc.2022.4564
[20] Rockafellar, R. T., & Uryasev, S. (2000). Optimization of conditional value at risk. Journal of Risk, 2(3), 21–41. DOI: https://doi.org/10.21314/JOR.2000.038
[21] Artzner, P., Delbaen, F., Eber, J. M., & Heath, D. (1999). Coherent measures of risk. Mathematical Finance, 9(3), 203–228. DOI: https://doi.org/10.1111/1467-9965.00068
[22] Gotoh, J., & Takano, Y. (2007). Newsvendor solutions via conditional value at risk minimization. European Journal of Operational Research, 179(1), 80–96. DOI: https://doi.org/10.1016/j.ejor.2006.03.022
[23] Chen, Y., Xu, M., & Zhang, Z. G. (2009). A risk averse newsvendor model under the CVaR criterion. Operations Research, 57(4), 1040–1044. DOI: https://doi.org/10.1287/opre.1080.0603
[24] Choi, S., Ruszczyński, A., & Zhao, Y. (2011). A multiproduct risk averse newsvendor with law invariant coherent measures of risk. Operations Research, 59(2), 346–364. DOI: https://doi.org/10.1287/opre.1100.0896
[25] Chen, X., Sim, M., Simchi Levi, D., & Sun, P. (2007). Risk aversion in inventory management. Operations Research, 55(5), 828–842. DOI: https://doi.org/10.1287/opre.1070.0429
[26] Wecker, W. E. (1978). Predicting demand from sales data in the presence of stockouts. Management Science, 24(10), 1043–1054. DOI: https://doi.org/10.1287/mnsc.24.10.1043
[27] Nahmias, S. (1994). Demand estimation in lost sales inventory systems. Naval Research Logistics, 41(6), 739–757. DOI: https://doi.org/10.1002/1520-6750(199410)41:6<739::AID-NAV3220410605>3.0.CO;2-A
[28] Agrawal, N., & Smith, S. A. (1996). Estimating negative binomial demand for retail inventory management with unobservable lost sales. Naval Research Logistics, 43(6), 839–861. DOI: https://doi.org/10.1002/(SICI)1520-6750(199609)43:6<839::AID-NAV4>3.0.CO;2-5
[29] Besbes, O., & Muharremoglu, A. (2013). On implications of demand censoring in the newsvendor problem. Management Science, 59(6), 1407–1424. DOI: https://doi.org/10.1287/mnsc.1120.1654
[30] Sachs, A. L., & Minner, S. (2014). The data driven newsvendor with censored demand observations. International Journal of Production Economics, 149, 28–36. DOI: https://doi.org/10.1016/j.ijpe.2013.04.039
[31] Ban, G. Y. (2020). Confidence intervals for data driven inventory policies with demand censoring. Operations Research, 68(2), 309–326. DOI: https://doi.org/10.1287/opre.2019.1883
[32] Paterson, C., Kiesmüller, G., Teunter, R., & Glazebrook, K. (2011). Inventory models with lateral transshipments: A review. European Journal of Operational Research, 210(2), 125–136. DOI: https://doi.org/10.1016/j.ejor.2010.05.048
[33] Griffin, E. C., Keskin, B. B., & Allaway, A. W. (2023). Clustering retail stores for inventory transshipment. European Journal of Operational Research, 311(2), 690–707. DOI: https://doi.org/10.1016/j.ejor.2023.06.008
[34] Shapiro, A., Dentcheva, D., & Ruszczyński, A. (2021). Lectures on stochastic programming: Modeling and theory (3rd ed.). SIAM. DOI: https://doi.org/10.1137/1.9781611976595
[35] Ma, S., Fildes, R., & Huang, T. (2016). Demand forecasting with high dimensional data: The case of SKU retail sales forecasting with intra and inter category promotional information. European Journal of Operational Research, 249(1), 245–257. DOI: https://doi.org/10.1016/j.ejor.2015.08.029
[36] Trapero, J. R., Kourentzes, N., & Fildes, R. (2015). On the identification of sales forecasting models in the presence of promotions. Journal of the Operational Research Society, 66(2), 299–307. DOI: https://doi.org/10.1057/jors.2013.174
[37] Cooper, L. G., Baron, P., Levy, W., Swisher, M., & Gogos, P. (1999). PromoCast: A new forecasting method for promotion planning. Marketing Science, 18(3), 301–316. DOI: https://doi.org/10.1287/mksc.18.3.301
[38] Fildes, R., Goodwin, P., & Önkal, D. (2019). Use and misuse of information in supply chain forecasting of promotion effects. International Journal of Forecasting, 35(1), 144–156. DOI: https://doi.org/10.1016/j.ijforecast.2017.12.006
[39] Salinas, D., Flunkert, V., Gasthaus, J., & Januschowski, T. (2020). DeepAR: Probabilistic forecasting with autoregressive recurrent networks. International Journal of Forecasting, 36(3), 1181–1191. DOI: https://doi.org/10.1016/j.ijforecast.2019.07.001
[40] Koenker, R., & Bassett, G. (1978). Regression quantiles. Econometrica, 46(1), 33–50. DOI: https://doi.org/10.2307/1913643
[41] Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T. Y. (2017). LightGBM: A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems 30 (pp. 3149–3157).
[42] Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3), 253–263. DOI: https://doi.org/10.1080/07350015.1995.10524599
[43] Chen, Z., Wang, M., Zeng, Z., & Ping, W. (2026). Uncertainty aware financial forecasting: Leveraging conformal prediction for risk adjusted models. IEEE Access, 14, 90053–90077. DOI: https://doi.org/10.1109/ACCESS.2026.3702666
[44] Zhang, H., Ge, Y., Zhao, X., & Wang, J. (2025). Hierarchical deep reinforcement learning for multi objective integrated circuit physical layout optimization with congestion aware reward shaping. IEEE Access, 13, 162533–162551. DOI: https://doi.org/10.1109/ACCESS.2025.3610615
[45] Teng, D. (2025). TEAS: Token and energy aware autoscaling for cost efficient LLM serving. AI Data Science Journal, 2(3), 47–53. DOI: https://doi.org/10.61784/adsj3036
[46] Wang, B., Wang, Z., Zhao, W., Zhang, F., & Shang, W. (2026). DRL Adapt: Deep reinforcement learning for adaptive routing convergence optimization in large scale networks. IEEE Open Journal of the Computer Society, 7, 849–863. DOI: https://doi.org/10.1109/OJCS.2026.3687441
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Highlights in Business, Economics and Management
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
