Using Operational Data Analytics for Planning Decisions Under Uncertainty

We consider a planning problem where events occur according to a renewal process with unknown parameters for the interevent times. We wish to make a decision to prepare for the n + Nth event from now. For that, we explore by using the first n events to obtain samples of the time between events and then exploit to make a decision related to the Nth event from that time. The traditional approach is to first estimate the parameters of the interevent time distribution using the n samples, and then make an optimal decision. However, it has been well documented that such an approach of estimation and then optimization is oftentimes suboptimal. To redress that concern, a framework called operational data analytics (ODA) has recently been garnering attention. To the best of our knowledge, ODA has not been adapted to the above decision-making setting, and we develop an analytical solution using ODA. Further, we perform extensive numerical experiments to illustrate (a) the difference in the solution quality between an oracle that knows the parameters and estimating them from the sample; (b) the improvement in using ODA’s effectiveness versus using traditional statistical estimation and then optimization, a Bayesian approach, and a bootstrapping method; (c) the asymptotic results used for verification; and (d) its effectiveness in a real-data setting where events do not necessarily follow a renewal process. History: Eunshin Byon served as the senior editor for this article. Data Ethics & Reproducibility Note: The code capsule is available on Code Ocean at https://codeocean.com/capsule/1158767/tree/v1 and in the Supplemental Materials to this article (available at https://doi.org/10.1287/ijds.2024.0051 ).