SCENARIO RECENCY BEFORE RISK AVERSION: MODULAR CAPACITY PLANNING UNDER NON-STATIONARY RETAIL ACTIVITY
Özet
The activity used for planning a future processing period is not known when capacity has to be installed. The usual formulation compares a decision that minimizes expected cost with one that also places weight on adverse outcomes. Such a comparison leaves an important matter unresolved: the probability distribution used for the calculation may describe an earlier operating level rather than the one for which capacity is needed. We investigate that matter with 277 weekly aggregates from the M5 retail sales data. The observations cover three departments at each of two stores. Capacity is represented by a bounded integer decision, while a single pooled resource is allocated exactly and divisibly after the workload becomes known. Exposure to adverse outcomes is measured using 90% conditional value at risk (CVaR), and historical scenarios are constructed in different ways. For each store, expanding estimation histories precede three successive, non-overlapping test periods of 24 weeks each. With 52 historical weeks selected at even intervals, mean held-out cost is 2.345 under risk neutrality. It falls to 1.952 with a fixed risk coefficient of 0.25; selecting risk preference through validation instead gives 2.002. A risk-neutral model using the latest 52 training weeks reaches 1.949, while a fixed coefficient of 0.25 on all available training weeks gives 1.948. These are different means of representing the decision, but neither recency nor risk aversion is shown to be universally preferable. A risk penalty can look more helpful when the historical scenarios are no longer well aligned with the operating period. The chapter explains the newsvendor-like structure of the simplified capacity problem, separates reductions in tail loss implied by that structure from evidence of decision performance, and examines the full sensitivity of validation-based calibration. Processing rates and cost coefficients are illustrative. The data are recorded sales, not uncensored demand, and the comparisons constitute controlled decision experiments rather than measurements of industrial savings.
Anahtar Kelimeler: capacity planning, scenario recency, non-stationarity, newsvendor, CVaR, decision-focused evaluation, M5 data
Konu Alanı: Mühendislik -> Mühendislik (diğer) -> Endüstri Mühendisliği
REFERENCES
Ahmed, S., King, A. J., & Parija, G. (2003). A multi-stage stochastic integer programming approach for capacity expansion under uncertainty. Journal of Global Optimization, 26, 3–24. https://doi.org/10.1023/A:1023062915106
Bertsimas, D., & Kallus, N. (2020). From predictive to prescriptive analytics. Management Science, 66(3), 1025–1044. https://doi.org/10.1287/mnsc.2018.3253
Delage, E., & Ye, Y. (2010). Distributionally robust optimization under moment uncertainty with application to data-driven problems. Operations Research, 58(3), 595–612. https://doi.org/10.1287/opre.1090.0741
Elmachtoub, A. N., & Grigas, P. (2022). Smart “predict, then optimize”. Management Science, 68(1), 9–26. https://doi.org/10.1287/mnsc.2020.3922
Eppen, G. D., Martin, R. K., & Schrage, L. (1989). OR practice—A scenario approach to capacity planning. Operations Research, 37(4), 517–527. https://doi.org/10.1287/opre.37.4.517
Kannan, R., Bayraksan, G., & Luedtke, J. R. (2025). Technical note—Data-driven sample average approximation with covariate information. Operations Research, 73(6), 3245–3259. https://doi.org/10.1287/opre.2020.0533
Khouja, M. (1999). The single-period (news-vendor) problem: Literature review and suggestions for future research. Omega, 27(5), 537–553. https://doi.org/10.1016/S0305-0483(99)00017-0
Luss, H. (1982). Operations research and capacity expansion problems: A survey. Operations Research, 30(5), 907–947. https://doi.org/10.1287/opre.30.5.907
Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2022a). M5 accuracy competition: Results, findings, and conclusions. International Journal of Forecasting, 38(4), 1346–1364. https://doi.org/10.1016/j.ijforecast.2021.11.013
Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2022b). The M5 competition: Background, organization, and implementation. International Journal of Forecasting, 38(4), 1325–1336. https://doi.org/10.1016/j.ijforecast.2021.07.007
Mohajerin Esfahani, P., & Kuhn, D. (2018). Data-driven distributionally robust optimization using the Wasserstein metric: Performance guarantees and tractable reformulations. Mathematical Programming, 171, 115–166. https://doi.org/10.1007/s10107-017-1172-1
Noyan, N. (2012). Risk-averse two-stage stochastic programming with an application to disaster management. Computers & Operations Research, 39(3), 541–559. https://doi.org/10.1016/j.cor.2011.03.017
Rockafellar, R. T., & Uryasev, S. (2000). Optimization of conditional value-at-risk. The Journal of Risk, 2(3), 21–41. https://doi.org/10.21314/JOR.2000.038
Rockafellar, R. T., & Uryasev, S. (2002). Conditional value-at-risk for general loss distributions. Journal of Banking & Finance, 26(7), 1443–1471. https://doi.org/10.1016/S0378-4266(02)00271-6
Van Mieghem, J. A. (2003). Capacity management, investment, and hedging: Review and recent developments. Manufacturing & Service Operations Management, 5(4), 269–302. https://doi.org/10.1287/msom.5.4.269.24882
Safaei, M. (2026). SCENARIO RECENCY BEFORE RISK AVERSION: MODULAR CAPACITY PLANNING UNDER NON-STATIONARY RETAIL ACTIVITY. In EROĞLU, Y. (Ed.), Optimization and Strategic Decision Models in Industrial Engineering (pp. -). Vizetek Yayıncılık. https://doi.org/10.54637/vizetek.9786253825010
| Kitap: |
| Yayınevi:
Vizetek Yayıncılık |