Knowledge that Transforms

The Importance of Considering System Congestion When Acquiring Used Items for Remanufacturing The condition of used items obtained by remanufacturers often varies widely and may only be truly known after an inspection. In “Technical Note–Optimal Procurement in Remanufacturing Systems with Uncertain Used-Item Condition,” Nadar, Akan, Debo, and Scheller-Wolf study the procurement problem by taking into account this pre-inspection uncertainty as well as the quantities of available used items of different quality levels. The authors provide a novel optimal policy structure for a Markov decision process representation of this problem. They prove that the optimal acquisition decision depends on the system congestion level — the total remanufacturing load weighted by the quality levels. Their results show that it becomes less desirable to acquire a used item as the number of available items or the quality level of any available item increases. To prove their results, the authors introduce a new functional characterization that is a relaxation of supermodularity and discrete convexity.

In the Bayesian newsvendor problem, it is known that the optimal decision is always greater than or equal to the myopic decision. As a result, the optimal decision can be expressed as the sum of the myopic decision plus a nonnegative “exploration boost.” In “Bayesian Inventory Control: Accelerated Demand Learning via Exploration Boosts,” Chuang and Kim characterize the form of the exploration boost in terms of basic statistical measures of uncertainty. This characterization expresses in clear terms the way in which the statistical learning and inventory control are jointly optimized; when there is a high degree of parameter uncertainty, inventory levels are boosted to induce a higher chance of observing more sales data to more quickly resolve statistical uncertainty, and as parameter uncertainty resolves, the exploration boost is reduced.

Crowds management during peak commuting hours is a key challenge facing metro systems worldwide, which results in serious safety concerns and unfair public transit service for commuters on different origin-destination (o-d) pairs. In “Online Passenger Flow Control in Metro Lines,” the authors investigate the impact of online decision making on the value of passenger flow control solution methodologies. The authors formulate the problem as a stochastic dynamic program with a fairness (fill rate) constraint and exploit Blackwell's approachability theorem and Fenchel duality to characterize the attainable service level of each o-d pair. They use these insights to develop online policies that can enable more passengers boarding a train (efficiency) as well as ensure equitable service level (fairness) provided to each o-d pair. Numerical experiments on a set of transit data from Beijing show that this approach performs well compared with existing benchmarks in the literature.

As a result of digitization of the economy, more and more decision makers from a wide range of domains have gained the ability to target products, services, and information provision based on individual characteristics. Examples include selecting offers, prices, advertisements, or emails to send to consumers, choosing a bid to submit in a contextual first-price auctions, and determining which medication to prescribe to a patient. The key to enabling this is to learn a treatment policy from historical observational data in a sample-efficient way, hence uncovering the best personalized treatment choice recommendation. In “Offline Policy Learning: Generalization and Optimization,” Z. Zhou, S. Athey, and S. Wager provide a sample-optimal policy learning algorithm that is computationally efficient and that learns a tree-based treatment policy from observational data. In our quest toward fully automated personalization, the work provides a theoretically sound and practically implementable approach.

In this paper, we introduce a new method for evaluating the performance of static pricing in one-way vehicle-sharing systems. Our approach, based on a well-known recursive relationship, leads to a series of increasingly tight bounds on the performance of the static pricing policy. These bounds are valid for systems with multiple locations, nonzero travel times, and an arbitrary number of vehicles. They also apply to systems where the static pricing policy does not lead to a fully connected network. Our method results in a family of asymptotically optimal static pricing policies that improve upon previous results in the literature. The approach applies to the case of a single location and yields a bound that is at least as tight as the best known bound.

Efficient Learning Algorithms for Dynamic Inventory Allocation in Multiwarehouse Multistore Systems with Censored Demand Motivated by collaboration with a prominent fast-fashion retailer in Europe, the researchers focus their attention on the one-warehouse multistore (OWMS) inventory control problem, specifically addressing scenarios in which the demand distribution is unknown a priori. The OWMS problem revolves around a central warehouse that receives initial replenishments and subsequently distributes inventory to multiple stores within a finite time horizon. The objective lies in minimizing the total expected cost. To overcome the hurdles posed by the unknown demand distribution, the researchers propose a primal-dual algorithm that continuously learns from demand observations and dynamically adjusts inventory control decisions in real time. Thorough theoretical analysis and empirical evaluations highlight the promising performance of this approach, offering valuable insights for efficient inventory allocation within the ever-evolving retail industry.

A Data-Driven Approach to Improve Care Unit Placements in Hospitals The choice of care unit upon hospital admission is a challenging task because of the wide variety of patient characteristics, uncertain needs of patients, and limited number of beds in intensive and intermediate care units. These decisions require carefully weighing the benefits of improved health outcomes against the opportunity cost of reserving higher level care beds for potentially more complex patients arriving in the future. In “Data-Driven Hospital Admission Control: A Learning Approach,” Zhalechian, Keyvanshokooh, Shi, and Van Oyen introduce a data-driven algorithm to address this challenging task. By focusing on reducing the readmission risk of patients, the algorithm is designed to (i) adaptively learn the readmission risk of patients through batch learning with delayed feedback and (ii) determine the best care unit placement for a patient based on the observed information and occupancy levels to minimize total readmission risk. The algorithm is supported by a performance guarantee, and its effectiveness is showcased using real-world hospital system data.

How Product Locations Drive Traffic Throughout a Retail Store In “Store-Wide Shelf-Space Allocation with Ripple Effects Driving Traffic,” Flamand, Ghoniem, and Maddah develop a framework for deciding where to place products in a store, in addition to apportioning the shelf space among products, in a way that maximizes impulse profit, a phenomenon that may account for 50% of transactions. By analyzing a large data set of customer receipts from a grocery store in Beirut, the authors develop a regression model that estimates traffic at a shelf based on its location and the “attraction” from products allocated nearby. The traffic model is embedded within a mixed-integer nonlinear program, which they solve via specialized linear approximations. For the store in Beirut, a 65% improvement in impulse profit is anticipated, and the location of products is found to be significantly more important in driving store-wide traffic than the relative shelf-space allocation.

Partially Recovering a Graph Alignment in the Correlated Erdös–Renyi Model Given two graphs, how can we partially recover a one-to-one mapping between nodes that maximizes edge overlap? This problem, known as the graph alignment problem, appears in settings such as social network deanonymization and cellular biology. In “Partial Recovery in the Graph Alignment Problem,” G. Hall and L. Massoulié consider a stylized mathematical model of problems of this type: they assume that the input graphs are generated via a probabilistic model, namely, the correlated Erdös–Renyi model with parameters (n, q, s). They provide both necessary and sufficient conditions on (n, q, s) under which partial recovery can be achieved. In particular, they show that partial recovery can be achieved in the nqs = Ɵ(1) regime under certain additional assumptions.

The recent coronavirus disease 2019 pandemic has shown that shortages and supply chain disruptions can have catastrophic effects on the real economy. These observations bring about reflections and first-order questions. How can we design supply chain networks that are robust and resilient to demand and supply shocks? Can we quantify the indirect effects caused by buyers’ and suppliers’ defaults in the network? Is it always cost effective to steer the system toward higher buyers’ and suppliers’ diversification? In the paper “Disruption and Rerouting in Supply Chain Networks,” Birge et al. argue that in highly capitalized networks, diversifying demand and supply across a larger number of counterparties may result in a more fragile network. Single-sourcing strategies are optimal for a firm only if the firm’s supplier default probability is low, but they perform worse than multiple-sourcing strategies otherwise.