This work provides performance guarantees for solving data-driven contextual newsvendor problems when the contextual data contains intertemporal dependence and non-stationarities. While machine learning tools have observed increasing use in data-driven inventory management problems, most of the existing work assumes that the contextual data are independent and identically distributed (often referred to as i.i.d.). However, such assumptions are often violated in real operational environments where the contextual data are sequentially generated with intertemporal correlations and possible non-stationarities. By accommodating these naturally arising operational environments, our work adopts comparatively more realistic assumptions and develops out-of-sample performance bounds for learning data-driven contextual newsvendor problems.
BesbesOMouchtakiO (2023) How big should your data really be? Data-driven newsvendor: Learning one sample at a time. Management Science69(10): 5848–5865.
8.
BeutelALMinnerS (2012) Safety stock planning under causal demand forecasting. International Journal of Production Economics140(2): 637–645.
9.
BousquetOElisseeffA (2002) Stability and generalization. Journal of Machine Learning Research2(Mar): 499–526.
10.
BrockwellPJDavisRA (eds) (2002) Introduction to time series and forecasting. New York, NY, USA: Springer New York.
11.
CaoYShenZJM (2019) Quantile forecasting and data-driven inventory management under nonstationary demand. Operations Research Letters47(6): 465–472.
12.
CarrizosaEOlivares-NadalAVRamírez-CoboP (2016) Robust newsvendor problem with autoregressive demand. Computers & Operations Research68: 123–133.
13.
ChuLYShanthikumarJGShenZJM (2008) Solving operational statistics via a Bayesian analysis. Operations Research Letters36(1): 110–116.
14.
DavydovYA (1973) Mixing conditions for Markov chains. Teoriya Veroyatnostei i ee Primeneniya18(2): 321–338.
15.
DontiPAmosBKolterJZ (2017) Task-based end-to-end model learning in stochastic optimization. Advances in Neural Information Processing Systems30.
16.
DoukhanP (2012) Mixing: properties and examples, volume 85. Springer Science & Business Media.
El BalghitiOElmachtoubANGrigasP, et al. (2019) Generalization bounds in the predict-then-optimize framework. Advances in Neural Information Processing Systems32. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc.
19.
ElmachtoubANGrigasP (2022) Smart “predict, then optimize”. Management Science68(1): 9–26.
20.
HangHSteinwartI (2014) Fast learning from α-mixing observations. Journal of Multivariate Analysis127: 184–199.
21.
IbragimovI (1975) Independent and stationary sequences of random variables. Groningen, Netherlands: Wolters, Noordhoff Pub.
22.
KeskinNBMinXSongJSJ (2021) The nonstationary newsvendor: Data-driven nonparametric learning. Available at SSRN 3866171.
23.
KontorovichLARamananK (2008) Concentration inequalities for dependent random variables via the martingale method. The Annals of Probability36(6): 2126–2158.
24.
KuznetsovVMohriM (2014) Generalization bounds for time series prediction with non-stationary processes. International conference on algorithmic learning theory, 260–274 (Springer).
KwiatkowskiDPhillipsPCSchmidtP, et al. (1992) Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root?Journal of Econometrics54(1–3): 159–178.
27.
LeviRPerakisGUichancoJ (2015) The data-driven newsvendor problem: New bounds and insights. Operations Research63(6): 1294–1306.
28.
LeviRRoundyROShmoysDB (2007) Provably near-optimal sampling-based policies for stochastic inventory control models. Mathematics of Operations Research32(4): 821–839.
29.
McdonaldDShaliziCSchervishM (2011a) Estimating beta-mixing coefficients. Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics. JMLR Workshop and Conference Proceedings, 516–524.
McDonaldDJShaliziCRSchervishM (2015) Estimating beta-mixing coefficients via histograms. Estimating Beta-Mixing Coefficients via Histograms9: 2855–2883.
32.
MitrophanovAY (2005) Sensitivity and convergence of uniformly ergodic Markov chains. Journal of Applied Probability42(4): 1003–1014.
33.
MohriMRostamizadehA (2010) Stability bounds for stationary φ-mixing and β-mixing processes. Journal of Machine Learning Research11(Feb): 789–814.
34.
NeghabDPKhayyatiSKaraesmenF (2022) An integrated data-driven method using deep learning for a newsvendor problem with unobservable features. European Journal of Operational Research302(2): 482–496.
35.
OroojlooyjadidASnyderLVTakáčM (2020) Applying deep learning to the newsvendor problem. IISE Transactions52(4): 444–463.
QiMShiYQiY, et al. (2020) A practical end-to-end inventory management model with deep learning. Available at SSRN 3737780.
38.
SamsonPM (2000) Concentration of measure inequalities for Markov chains and -mixing processes. The Annals of Probability28(1): 416–461.
39.
SteinwartIChristmannA (2011) Estimating conditional quantiles with the help of the pinball loss. Bernoulli17(1): 211–225.
40.
SteinwartIHushDRScovelC, et al. (2009) Optimal rates for regularized least squares regression. COLT: 79–93.
41.
WahbaG (2000) An introduction to model building with reproducing kernel Hilbert spaces. Statistics Department TR 1020.
42.
ZouBLiLXuZ (2009) The generalization performance of ERM algorithm with strongly mixing observations. Machine Learning75(3): 275–295.
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
