论文速递 | European Journal of Operational Research 9月文章精选

Chinese universities have experienced a significant wave of mergers since the late 1990s, raising questions about the impact of such consolidation on the knowledge output efficiency and economies of scale and scope in higher education. In this study, we propose a nonparametric model that takes into account specific output preferences during a preliminary evaluation of merger proposals. Additionally, our model addresses the challenges related to the existence of infeasible solutions that have been highlighted in related literature. Empirical analysis using a sample of Chinese research universities suggests that the current number of universities is below optimal in most provinces regardless of the type of knowledge outcome used to assess university performance. This suggests that further university mergers are unnecessary in these regions. However, multiple beneficial merger scenarios still exist in the provinces of Guizhou and Gansu. In Guizhou, some merger plans could lead to efficiency gains when scholarly publications are used as the knowledge outcome, while in Gansu, several consolidation scenarios show potential gains when research projects serve as outputs. Moreover, we found evidence of widespread diseconomies of scale across most merger scenarios along with relatively modest economies of scope within these two provinces.

自 20 世纪 90 年代末以来，中国大学经历了一波大规模的合并浪潮，这引发了关于此类整合对高等教育知识产出效率以及规模经济和范围经济影响的疑问 。本研究中，我们提出了一个非参数模型，在对合并提案进行初步评估时考虑了特定的产出偏好 。此外，我们的模型解决了相关文献中强调的与不可行解存在相关的挑战 。使用中国研究型大学样本的实证分析表明，无论使用何种知识产出类型来评估大学绩效，大多数省份当前的大学数量都低于最优水平 。这表明在这些地区进一步的大学合并是不必要的 。然而，在贵州和甘肃两省，仍然存在多种有益的合并情景 。在贵州，当以学术出版物作为知识产出时，一些合并计划可能带来效率提升；而在甘肃，当以科研项目作为产出时，若干整合情景显示出潜在收益 。此外，我们发现在大多数合并情景中存在普遍的规模不经济现象，同时在这两个省份内范围经济效应相对有限 。

Timely response in the aftermath of a disaster is crucial to alleviate loss of life and suffering. Timeliness of relief may be hampered by road network disruptions caused by the disaster, such as damage to road segments or debris covering the roads. The use of drones simultaneously with trucks can potentially help overcome issues around network disruptions and achieve more timely delivery of post-disaster aid. In an effort to shed more light into this potential, we address the problem of network design for parallel truck–drone operations by depot location prior to the disaster and routing of the vehicles in its aftermath. We incorporate the uncertainty on network disruption by modelling this problem as a two-stage stochastic program, which proves computationally challenging to solve to optimality for real-life disaster scenarios. Consequently, we propose a tailored heuristic based on variable neighbourhood search to find high-quality solutions efficiently. Our computational results on randomly generated instances and a case study from the 2011 Van Earthquake in Turkiye demonstrate the effectiveness of the heuristic, the benefits of employing both trucks and drones, and the significance of accounting for uncertainties in pre-disaster planning.

灾后及时响应对于减少生命损失和痛苦至关重要 。由于灾难导致的路网中断，如路段损坏或道路被碎石覆盖，可能会妨碍救援的及时性 。将无人机与卡车同时使用，可能有助于克服网络中断问题，实现更及时的灾后援助物资运送 。为深入探讨这一潜力，我们研究了卡车-无人机并行作业的网络设计问题，包括灾前仓库选址和灾后车辆路径规划 。我们通过将此问题建模为一个两阶段随机规划来纳入网络中断的不确定性，但对于现实生活中的灾难情景，求得最优解在计算上具有挑战性 。因此，我们提出了一种基于变邻域搜索的定制启发式算法，以高效地找到高质量解 。我们在随机生成的实例和 2011 年土耳其凡城地震的案例研究上的计算结果，证明了该启发式算法的有效性、同时使用卡车和无人机的益处，以及在灾前规划中考虑不确定性的重要性 。

The Distributed Flexible Job Shop Scheduling Problem (DFJSP) is a well-known -hard optimization problem with widespread applications in production scheduling. It involves assigning jobs to factories, allocating operations to machines, and sequencing operations on each machine, which presents significant computational challenges. Although heuristic and metaheuristic approaches have been extensively studied, the exploration of exact algorithms for solving DFJSP remains limited. This paper addresses this gap by proposing three logic-based Benders decomposition (LBBD) frameworks specifically designed for the DFJSP, leveraging the problem’s decomposable structure to achieve optimal or near-optimal solutions with quantifiable quality guarantees within strict time limits. In each LBBD framework, the DFJSP is decomposed into a master problem and several subproblems based on specific decomposition schemes. The corresponding Mixed-Integer Linear Programming (MILP) models and Constraint programming (CP) models for these problems are formulated and solved alternately. Additionally, a hybrid optimization approach is developed by integrating LBBD with CP and heuristic search strategies. The proposed method includes an enhanced CP model with targeted improvements to boost its solving efficiency and incorporates a critical path-based local search strategy to further refine the solution quality. Moreover, several strong subproblem relaxation schemes are incorporated into the master problem under different LBBD frameworks. Comprehensive evaluations on an extended benchmark dataset containing 286 instances demonstrate that the hybrid algorithm achieves an average optimality gap of less than 1.2%. Compared to state-of-the-art MILP, CP, and heuristic methods, the proposed approach delivers superior solution quality and computational efficiency, establishing a new benchmark for solving the DFJSP.

分布式柔性作业车间调度问题 (DFJSP) 是一个著名的 NP-hard 优化问题，在生产调度领域有广泛应用 。该问题涉及将作业分配给工厂、将工序分配给机器以及对每台机器上的工序进行排序，这带来了巨大的计算挑战 。尽管启发式和元启发式方法已被广泛研究，但求解 DFJSP 的精确算法的探索仍然有限 。本文通过提出三种专为 DFJSP 设计的基于逻辑的 Benders 分解 (LBBD) 框架来填补这一空白，利用该问题的可分解结构，在严格的时间限制内获得具有可量化质量保证的最优或近优解 。在每个 LBBD 框架中，DFJSP 根据特定的分解方案被分解为一个主问题和若干子问题 。我们为这些问题构建了相应的混合整数线性规划 (MILP) 模型和约束规划 (CP) 模型，并交替求解 。此外，通过将 LBBD 与 CP 和启发式搜索策略相结合，开发了一种混合优化方法 。所提出的方法包括一个经过针对性改进以提升求解效率的增强 CP 模型，并结合了基于关键路径的局部搜索策略以进一步改善解的质量 。此外，在不同的 LBBD 框架下，我们将几种强子问题松弛方案整合到主问题中 。在一个包含 286 个实例的扩展基准数据集上的综合评估表明，该混合算法实现了低于 1.2% 的平均最优性差距 。与最先进的 MILP、CP 和启发式方法相比，所提出的方法在求解质量和计算效率上均表现更优，为解决 DFJSP 问题建立了新的基准 。

● 题目：Integrated vessel traffic scheduling and berth allocation with restricted channel widths

考虑航道宽度限制的船舶交通调度与泊位分配一体化问题

● 主题：Production, Manufacturing, Transportation and Logistics

● 作者：

Hongda Duan (a) (b), Lixin Miao (b) (c), Shuai Jia (d) *, Canrong Zhang (a) (b), Jasmine Siu Lee Lam (e)

(a)Department of Industrial Engineering, Tsinghua University, Beijing 100084, China

(b)Research Center for Modern Logistics, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China

(c)Tsinghua-Berkeley Shenzhen Institute, Tsinghua University, Shenzhen 518055, China

(d)Thrust of Intelligent Transportation, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou 511400, China

(e)Department of Technology, Management and Economics, Technical University of Denmark, Denmark

● 发表时间：2025-09-15

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.016

● 重点：

• The vessel traffic scheduling problem is studied by considering channel width restrictions.
• A binary integer programming model is proposed with time- and space-discretization.
• A novel machine-learning-enhanced column generation method is proposed.
• Effectiveness of the method is evaluated on both synthetic and real-world cases.
• 研究了考虑航道宽度限制的船舶交通调度问题 。
• 提出了一个基于时间和空间离散化的二元整数规划模型 。
• 提出了一种新颖的机器学习增强的列生成方法 。
• 在合成案例和真实世界案例中评估了该方法的有效性 。

● 摘要：

Due to the surging volume of seaborne trade and high frequencies of port calls by vessels, seaports worldwide have been experiencing various levels of traffic congestion in the past few years. The incoming and outgoing vessel traffic in port areas is bottlenecked by the traffic infrastructure (e.g., navigation channels and inner anchorages) and the hydrological conditions (e.g., tidal effects) of a port, which can lead to significant congestion when vessel traffic and vessel service are not effectively planned. In this study, we investigate an integrated vessel traffic scheduling and berth allocation problem for congestion mitigation in a port. The problem encompasses the decision-making process of scheduling incoming and outgoing vessels in the navigation channels and anchorage areas of a port, and allocating berth space to vessels for service, so as to minimize the overall berthing and departure delay of vessels. In particular, we consider a practical scenario where the width of each channel may vary at different segments. This channel width restriction can render the problem much more difficult compared to a traditional setting with identical channel widths. We develop a binary integer programming model for the problem, and present a novel machine-learning-enhanced column generation algorithm for addressing this complex problem. Our method applies machine learning models to restrict vessels’ port stay times within limited time ranges, so that the search space for column generation can be reduced, leading to a trade-off between solution quality and computation efficiency. We validate the effectiveness of the proposed solution method utilizing both a case study of the Port of Shanghai and a computational study on synthetic large-scale instances.

由于海上贸易量激增和船舶挂靠港口频率的提高，全球海港在过去几年中经历了不同程度的交通拥堵 。港区的进出港船舶交通受到港口的交通基础设施（如航道和内部锚地）和水文条件（如潮汐效应）的瓶颈制约，如果船舶交通和船舶服务没有得到有效规划，可能导致严重拥堵 。本研究旨在探讨一个集成的船舶交通调度和泊位分配问题，以缓解港口拥堵 。该问题涵盖了在港口航道和锚地调度进出港船舶，并为船舶分配泊位进行服务的决策过程，目标是最小化船舶的总靠泊和离港延迟 。我们特别考虑了一个实际情景，即每个航道的宽度在不同段可能有所不同 。与传统航道宽度相同的设定相比，这种航道宽度限制会使问题变得更加困难 。我们为该问题建立了一个二元整数规划模型，并提出了一种新颖的机器学习增强的列生成算法来解决这个复杂问题 。我们的方法应用机器学习模型将船舶的港内停留时间限制在有限的时间范围内，从而可以缩小列生成的搜索空间，进而在解的质量和计算效率之间取得平衡 。我们通过上海港的案例研究和对合成大规模实例的计算研究，验证了所提出求解方法的有效性 。

文章5

● 题目：The optimal repair policy for an unreliable production system with limited available spare parts

备件有限的不可靠生产系统的最优维修策略

● 主题：Production, Manufacturing, Transportation and Logistics

● 作者：

Ludwig Brieditis (a) *, Gudrun P. Kiesmüller (a) (b), Filip Malmberg (a)

(a)Lund University, LTH Faculty of Engineering, Department of Industrial and Mechanical Sciences, Division of Production Management, P.O. Box 118, Lund, SE-221 00, Sweden

(b)Technical University of Munich, TUM School of Management, TUM Campus Heilbronn, Bildungscampus 9, 74076 Heilbronn, Germany

● 发表时间：2025-09-21

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.007

● 重点：

• We consider the repair problem for flow lines with shared spare part inventories.

• We formulate the problem as a Markov decision process.

• We conjecture the structure of optimal repair policies which may reserve spare parts.

• Numerical experiments suggest good performance of policies without reservations.

• 我们考虑了具有共享备件库存的流水生产线的维修问题 。

• 我们将该问题建模为一个马尔可夫决策过程 。

• 我们推测了可能保留备件的最优维修策略的结构 。

• 数值实验表明，不进行备件保留的策略表现良好 。

● 摘要：

In this study, a serial production system consisting of two machines and an intermediate finite buffer is considered. The machines have random processing times, and each machine contains one unit of a common critical component that is subject to random breakdown. Broken components must be replaced from an inventory of ready-for-use spare parts to restore machine functionality. This inventory is replenished according to a one-for-one replenishment policy with an externally given base stock level. Due to the limited availability of spare parts, the sequence in which the machines are to be repaired and whether they should be repaired immediately must be decided. The objective of our study is the maximization of the expected total discounted revenue per time unit over an infinite planning horizon.

We model the system as a semi-Markov decision process and characterize the optimal stationary repair policy. We show that the First-Break-First-Repair policy is not optimal and provide numerical evidence that the optimal repair decision depends on the number of units in the buffer and the number of available spare parts. We show that, in some system states, it is optimal to postpone repair of a machine and reserve spare parts for the other machine.

Since the optimal repair policy is state dependent and quite complex, we investigate different prioritization heuristics from the literature in a numerical study. Our experiments suggest that heuristics achieve excellent performance in practical settings.

本研究考虑一个由两台机器和一个中间有限缓冲区组成的串行生产系统 。机器的加工时间是随机的，每台机器包含一个会随机发生故障的通用关键部件 。损坏的部件必须从一个即用备件库存中更换，以恢复机器功能 。该库存根据“一对一”补货策略进行补充，其基础库存水平由外部给定 。由于备件供应有限，必须决定机器维修的顺序以及是否应立即进行维修 。我们研究的目标是在无限规划期内最大化单位时间的期望总折扣收益 。

我们将该系统建模为一个半马尔可夫决策过程，并刻画了最优平稳维修策略的特征 。我们证明了“先坏先修”策略并非最优，并提供了数值证据表明最优维修决策取决于缓冲区中的单元数量和可用备件的数量 。我们还表明，在某些系统状态下，推迟维修一台机器并为另一台机器保留备件是最优的 。

由于最优维修策略是状态依赖且相当复杂，我们在一个数值研究中考察了文献中的不同优先级启发式算法 。我们的实验表明，启发式算法在实际应用中表现出色 。

文章6

● 题目：Multi-cycle production network design under supply uncertainty

供应不确定性下的多周期生产网络设计

● 主题：Production, Manufacturing, Transportation and Logistics

● 作者：

Nadia Jaoui (a), Walid Klibi (b) *, Nizar El Hachemi (a) (c), Tarik Aouam (a) (d), Michel Fender (a)

(a)Africa Business School, Mohammed VI Polytechnic University, Rabat, 11103, Morocco

(b)Centre of Excellence for Supply Chain Innovation & Transportation (CESIT), Kedge Business School, Bordeaux, 33400, France

(c)Mohammadia School of Engineers, Mohammed V University, Rabat, 10090, Morocco

(d)Faculty of Economics and Business Administration, Ghent, 9000, Belgium

● 发表时间：2025-09-23

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.008

● 重点：

• Addressing a stochastic multi-cycle supply chain network design problem.

• Integrating supplier selection, facility location and capacity, and fleet sizing.

• Modeling uncertainty in supply, transportation, and production capacities.

• Proposing a new scenario-based partial Benders decomposition approach.

• Offering managerial insights based on a real-world supply chain case study.

• 解决一个随机性多周期供应链网络设计问题 。

• 整合了供应商选择、设施选址与产能以及车队规模确定 。

• 对供应、运输和生产能力的不确定性进行建模 。

• 提出了一种新的基于情景的部分 Benders 分解方法 。

• 基于一个真实世界的供应链案例研究提供管理见解 。

● 摘要：

This study addresses a novel production network design problem with an expanded scope inspired by a real-world business context. The problem involves strategic decisions for a long-term horizon regarding production sites’ location and capacity, suppliers’ selection, and transportation modes choice, given in-house vs external service providers’ options. To support these decisions, we integrate key tactical decisions for a large set of planning periods, involving flows between origin–destination pairs, production levels based on bill of materials, and inventory levels. Additionally, we consider uncertainty in raw material availability at suppliers, disruption in production capacities, and perturbation in transportation flows. First, we develop a multi-stage stochastic program that re-optimizes strategic decisions at each design period. Then, this program is reformulated into a multi-cycle two-stage stochastic model. Uncertainty is modeled through a finite set of scenarios generated using the Latin hypercube sampling technique, and the sample average approximation method is used to calibrate the sample size. Given the challenging solvability of the model, we proposed an advanced solution approach that builds on the recently introduced Partial Benders Decomposition (PBD) technique with new scenario creation strategies. Our experiments highlight the superiority of the proposed PBD’s variant in terms of solution quality and time reaching the best solution, compared to classical approaches. Furthermore, we demonstrate the benefits of enlarging the scope of the production network design problem by integrating all strategic decisions, which can yield gains of up to 36% compared to addressing them separately. Finally, we underscore the importance of stochastic modeling, contributing to cost reductions of over 3% compared to the deterministic counterpart.

本研究解决了一个受真实商业环境启发、范围更广的新型生产网络设计问题 。该问题涉及长期范围内的战略决策，包括生产地点的选址和产能、供应商的选择以及运输方式的选择，同时考虑了内部与外部服务提供商的选项 。为了支持这些决策，我们整合了大量规划周期内的关键战术决策，涉及起讫点之间的流量、基于物料清单的生产水平以及库存水平 。此外，我们考虑了供应商处原材料供应的不确定性、生产能力的中断以及运输流量的扰动 。首先，我们开发了一个多阶段随机规划，该规划在每个设计周期重新优化战略决策 。然后，该规划被重构为一个多周期两阶段随机模型 。不确定性通过使用拉丁超立方抽样技术生成的有限情景集进行建模，并使用样本均值近似法来校准样本大小 。鉴于模型的求解难度，我们提出了一种先进的求解方法，该方法基于最近引入的部分 Benders 分解 (PBD) 技术，并采用了新的情景创建策略 。我们的实验表明，与传统方法相比，所提出的 PBD 变体在解的质量和达到最佳解的时间方面均表现出优越性 。此外，我们证明了通过整合所有战略决策来扩大生产网络设计问题范围的好处，与分开处理相比，可带来高达 36% 的收益 。最后，我们强调了随机建模的重要性，与确定性模型相比，随机模型有助于降低超过 3% 的成本 。

文章7

● 题目：Markov decision processes: Monotonicity of optimal policy in exponential and quasi-hyperbolic discounting parameters

马尔可夫决策过程：最优策略在指数和拟双曲贴现参数下的单调性

● 主题：Stochastics and Statistics

● 作者：

Hakan Kılıç (a) *, Pelin Gülşah Canbolat (b), Evrim Didem Güneş (c)

(a)Institute of Communication, Culture, Information & Technology, University of Toronto Mississauga, 3359 Mississauga Rd, Mississauga, L5L 1C6, Canada

(b)Faculty of Engineering and Natural Sciences, Kadir Has University, Cibali Mah. Kadir Has Cad., Fatih, Istanbul, 34083, Turkiye

(c)College of Administrative Sciences and Economics, Koç University, Rumelifeneri Yolu, Sarıyer, Istanbul, 34450, Turkiye

● 发表时间：2025-09-23

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.013

● 重点：

• Conditions for monotonicity in quasi-hyperbolic discounting parameters are presented.

• Results cover general submodular or supermodular maximization problems.

• Applications to machine maintenance, patient behavior and inventory control are shown.

• 提出了拟双曲贴现参数下单调性的条件 。

• 研究结果涵盖了一般的子模或超模最大化问题 。

• 展示了在机器维护、患者行为和库存控制方面的应用 。

● 摘要：

Intertemporal preferences of decision makers, i.e., the way they discount delayed utilities, impact their decisions. Empirical evidence suggests that individuals commonly have hyperbolic discounting preferences. This can result in time-inconsistent behavior, e.g., procrastination, which may be a barrier to adopting preventive behavior such as machine maintenance and patient adherence to treatment. In this paper, we theoretically compare the actions of individuals based on their discounting characteristics. We consider the Hyperbolic Discounting (HD) model, which is more representative of individual behavior than Exponential Discounting (ED). We formulate a discrete-time finite-horizon Markov decision process with Quasi-Hyperbolic Discounting (QHD), an analytically tractable function representing HD and present sufficient conditions that ensure the monotonicity of the optimal policy in the discounting parameters. We consider submodular maximization or supermodular maximization problems. Our paper is the first to investigate the monotonicity of the optimal policy in QHD parameters for these problems. Moreover, we compare the optimal actions under ED and QHD. We apply our results to the settings of machine maintenance, individual health behavior and inventory control. We provide numerical examples that show there might not be monotonicity if our sufficient conditions are not met. Also, we explore the discrepancy between the expected total exponentially-discounted rewards of the actions obtained from QHD and of the actions that are optimal under ED, and observe that this discrepancy is affected mainly by the present bias.

决策者的跨期偏好，即他们对延迟效用的贴现方式，会影响他们的决策 。实证证据表明，个体普遍具有双曲贴现偏好 。这可能导致时间不一致的行为，例如拖延，这可能是采取预防性行为（如机器维护和患者坚持治疗）的障碍 。在本文中，我们从理论上根据个体的贴现特征比较他们的行为 。我们考虑了比指数贴现 (ED) 更能代表个体行为的双曲贴现 (HD) 模型 。我们构建了一个离散时间有限期马尔可夫决策过程，采用拟双曲贴现 (QHD)——一种能代表 HD 且易于分析的函数，并提出了确保最优策略在贴现参数上单调的充分条件 。我们考虑了子模最大化或超模最大化问题。我们的论文首次研究了这些问题中，最优策略在 QHD 参数下的单调性 。此外，我们比较了在 ED 和 QHD 下的最优行动 。我们将我们的结果应用于机器维护、个体健康行为和库存控制等情景 。我们提供了数值例子，表明如果我们的充分条件不满足，可能不存在单调性 。同时，我们探讨了从 QHD 获得的行动的期望总指数贴现收益与在 ED 下最优的行动的收益之间的差异，并观察到这种差异主要受当前偏误的影响 。

文章8

● 题目：Integrated estimate-and-optimize decision trees learning for two-stage linear decision-making problems

用于两阶段线性决策问题的“估计与优化”一体化决策树学习

● 主题：Analytics, Computational Intelligence and Information Management

● 作者：

Rafaela Ribeiro, Bruno Fanzeres *

Industrial Engineering Department, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), 22451-900, Rio de Janeiro, RJ, Brazil

● 发表时间：2025-09-24

● 原文链接: https://doi.org/10.1016/j.ejor.2025.08.048

● 重点：

• New framework integrates prediction and prescription for better decisions.

• Focus on decision trees and two-stage stochastic problems.

• Reformulates the proposed non-convex conceptual model as a MIP problem.

• Proposes a scalable Heuristic strategy leveraging decision tree structure.

• Enables fast decisions for a new context via precomputed, leaf-based prescriptions.

• 新框架整合了预测与决策以实现更优的决策 。

• 聚焦于决策树和两阶段随机问题 。

• 将所提出的非凸概念模型重构为一个混合整数规划 (MIP) 问题 。

• 提出了一种利用决策树结构的可扩展启发式策略 。

• 通过预先计算的、基于叶节点的决策方案，实现对新情景的快速决策 。

● 摘要：

Several decision-making under uncertainty problems found in industry and the scientific community can be framed as stochastic programs. Traditionally, these problems are addressed using a sequential two-step process, referred to as predict/estimate-then-optimize, in which a predictive distribution of the uncertain parameters is firstly estimated and then used to prescribe a decision. However, most predictive methods focus on minimizing forecast error, without accounting for its impact on decision quality. Moreover, practitioners often emphasize that their main goal is to obtain near-optimal solutions with minimum decision error, rather than least-error predictions. Therefore, in this work, we discuss a new framework for integrating prediction and prescription into the predictive distribution estimation process to be subsequently used to devise a decision. We particularly focus on decision trees and study decision-making problems representable as contextual two-stage linear programs. Firstly, we propose a workable framework along with a non-convex optimization model to account for the impact of the underlying decision-making problem on the predictive distribution estimation process. Then, we recast the non-convex model as a Mixed-Integer Programming (MIP) problem. Acknowledging the difficulty of the MIP reformulation to scale to large-scale instances, we devise a computationally efficient Heuristic strategy for the estimation problem leveraging the structure intrinsic to decision trees. A key feature of the proposed decision-making framework is its ability to instantly assess decisions by mapping new contexts to a leaf and retrieving the precomputed solution of the corresponding two-stage problem. A set of numerical experiments is conducted to illustrate the capability and effectiveness of the proposed framework using three distinct two-stage decision-making problems. We benchmark the proposed approach against prescriptions devised by various alternative frameworks. Five predict/estimate-then-optimize benchmarks that rely on commonly used predictive and distribution estimation methods and three benchmarks based on integrated predict-and-optimize decision-making processes are considered. We focus on evaluating solution quality and the computational performance of the MIP reformulation.

在工业界和科学界发现的许多不确定性下的决策问题都可以被构建为随机规划 。传统上，这些问题通过一个顺序的两步过程来解决，即“预测/估计-然后-优化”，其中首先估计不确定参数的预测分布，然后用它来制定决策 。然而，大多数预测方法侧重于最小化预测误差，而没有考虑其对决策质量的影响 。此外，实践者通常强调，他们的主要目标是获得决策误差最小的近优解，而不是误差最小的预测 。因此，在这项工作中，我们讨论了一个新的框架，将预测和决策整合到预测分布的估计过程中，以便随后用于制定决策 。我们特别关注决策树，并研究可表示为情景两阶段线性规划的决策问题 。首先，我们提出了一个可行的框架以及一个非凸优化模型，以考虑潜在决策问题对预测分布估计过程的影响 。然后，我们将该非凸模型重构为一个混合整数规划 (MIP) 问题 。考虑到 MIP 重构难以扩展到大规模实例，我们利用决策树的内在结构为该估计问题设计了一种计算高效的启发式策略 。所提出的决策框架的一个关键特征是，它能够通过将新情景映射到一个叶节点并检索相应两阶段问题的预计算解，来即时评估决策 。我们进行了一系列数值实验，使用三个不同的两阶段决策问题来说明所提出框架的能力和有效性 。我们将所提出的方法与各种替代框架制定的决策进行了基准比较 。我们考虑了五个依赖于常用预测和分布估计方法的“预测/估计-然后-优化”基准，以及三个基于一体化“预测并优化”决策过程的基准 。我们重点评估了 MIP 重构的解质量和计算性能 。

文章9

● 题目：Enhancing operational decision-making in fault diagnosis for high-dimensional data streams with auxiliary information

结合辅助信息增强高维数据流故障诊断中的运营决策

● 主题：Stochastics and Statistics

● 作者：

Zhihan Zhang (a) (1), Wendong Li (a) (1), Min Xie (b), Dongdong Xiang (a) *

(a)KLATASDS-MOE, School of Statistics, East China Normal University, Shanghai, China

(b)Department of Systems Engineering and Engineering Management, City University of Hong Kong, Kowloon, Hong Kong

● 发表时间：2025-09-25

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.022

● 重点：

• Post-signal fault diagnosis for high-dimensional data streams in modern engineering systems.

• Enhancing operational decision-making in fault diagnosis by integrating the auxiliary information.

• Cartesian hidden Markov model to describe sequential interdependencies.

• Minimizing the expected number of false positives while maintaining control over the missed discovery rate.

• 现代工程系统中高维数据流的信号后故障诊断 。

• 通过整合辅助信息来增强故障诊断中的运营决策 。
  

• 采用笛卡尔隐马尔可夫模型来描述序列间的相互依赖关系 。

• 在控制漏发现率的同时，最小化期望的误报数量 。

● 摘要：

Modern engineering systems, from advanced manufacturing processes to sophisticated electronic devices, generate high-dimensional data streams (HDS) that demand efficient operational strategies for quality management. While real-time anomaly detection is crucial, the importance of accurate post-signal fault diagnosis for root cause analysis has grown substantially. Current diagnostic methods often focus on isolated sequences of HDS, missing opportunities to leverage auxiliary information that can enhance decision-making. This paper introduces a novel framework to improve large-scale fault diagnosis in HDS environments, integrating auxiliary sequences within a multi-sequence multiple testing framework. Utilizing a Cartesian hidden Markov model, we develop a generalized local index of significance (GLIS) to assess the abnormality likelihood across data streams. Based on the GLIS, our proposed data-driven diagnostic procedure effectively harnesses auxiliary information, aiming to optimize operational decisions by minimizing the expected number of false positives in the primary sequence while maintaining control over the missed discovery rate. The asymptotic validity and optimality of this approach ensure its robustness in practical settings. We validate the efficacy of our method through comprehensive simulations and a real-world case study, demonstrating its potential to support more accurate and informed operational decisions.

现代工程系统，从先进的制造过程到复杂的电子设备，都会产生高维数据流 (HDS)，需要高效的运营策略进行质量管理 。虽然实时异常检测至关重要，但为进行根本原因分析而进行的精确信号后故障诊断的重要性已显著增加 。当前的诊断方法通常关注孤立的 HDS 序列，忽略了利用可增强决策的辅助信息的机会 。本文引入了一个新颖的框架，以改善 HDS 环境中的大规模故障诊断，该框架在一个多序列多重检验框架内整合了辅助序列 。利用笛卡尔隐马尔可夫模型，我们开发了一个广义局部显著性指数 (GLIS) 来评估跨数据流的异常可能性 。基于 GLIS，我们提出的数据驱动诊断程序有效地利用了辅助信息，旨在通过在控制漏发现率的同时最小化主序列中期望的误报数量来优化运营决策 。该方法的渐近有效性和最优性确保了其在实际应用中的稳健性 。我们通过全面的模拟和真实案例研究验证了我们方法的有效性，展示了其支持更准确、更明智的运营决策的潜力 。

文章10

● 题目：Distributionally robust optimization with generalized total variation ambiguity sets

基于广义总变差模糊集的分布鲁棒优化

● 主题：Stochastics and Statistics

● 作者：

Belleh Fontem (a) *, Ran Ji (b)

(a)Manning School of Business, University of Massachusetts Lowell, United States of America

(b)Department of Systems Engineering and Operations Research, George Mason University, United States of America

● 发表时间：2025-09-27

● 原文链接: https://doi.org/10.1016/j.ejor.2025.09.024

● 重点：

• Introduces GTV ambiguity sets adapting to sample size, support, and cost structure.

• Proposes two tractable methods: first-order and gradient-based, with guarantees.

• Gradient-based method exploits interior geometry of the ambiguity set.

• Validated on synthetic Newsvendor problems and real S&P 500 portfolio data.

• Outperforms Wasserstein and total variation sets in certain small-sample settings.

• 引入了能适应样本量、支撑集和成本结构的广义总变差 (GTV) 模糊集 。

• 提出了两种带保证的可行方法：一阶法和基于梯度的方法 。

• 基于梯度的方法利用了模糊集的内部几何结构 。

• 在合成的报童问题和真实的标普 500 投资组合数据上进行了验证 。

• 在某些小样本设置中优于 Wasserstein 模糊集和总变差模糊集 。

● 摘要：

This paper introduces a data-driven framework for distributionally robust optimization (DRO), founded on a new class of ambiguity sets termed generalized total variation (GTV) sets. In contrast to traditional DRO approaches, the proposed scheme constructs ambiguity sets whose geometry incorporates sample size, support, confidence level, empirical distribution, and cost function structure. Under this framework, we develop two tractable solution methods (first-order, gradient-based), each offering finite-sample statistical guarantees. The first-order approach employs sequential convex programming to construct a solution, followed by a linear program to determine a high-confidence upper bound (i.e., generalization bound) on the solution’s unknown true risk. The gradient-based approach, applicable when the ambiguity set has a smoothly curved boundary, utilizes gradient information to establish a high-confidence upper bound through a sequence of convex programs, all linear except for the final step. We prove that both methods produce statistically consistent risk estimates. Then, we empirically validate the framework on two applications: a synthetic two-item Newsvendor problem and a real-world portfolio optimization problem using S&P 500 asset returns. Results demonstrate that for finite support problems, GTV ambiguity sets can deliver generalization bounds that are as tight as, or tighter than, those from popular alternatives such as Wasserstein and total variation ambiguity sets. We thus highlight the practical benefits of incorporating several types of information into ambiguity set construction, offering improved robustness-performance tradeoffs for data-driven decision-making under uncertainty.

本文介绍了一种基于一类新的名为广义总变差 (GTV) 模糊集的数据驱动的分布鲁棒优化 (DRO) 框架 。与传统的 DRO 方法相比，该方案构建的模糊集的几何结构融合了样本量、支撑集、置信水平、经验分布和成本函数结构 。在此框架下，我们开发了两种可行的求解方法（一阶法和基于梯度的方法），每种方法都提供有限样本的统计保证 。一阶法采用序贯凸规划来构建解，然后通过一个线性规划来确定解的未知真实风险的一个高置信度上界（即泛化界） 。当模糊集具有平滑弯曲边界时，基于梯度的方法适用，它利用梯度信息通过一系列凸规划（除最后一步外均为线性）来建立一个高置信度上界 。我们证明了两种方法都能产生统计上一致的风险估计 。然后，我们在两个应用上对该框架进行了实证验证：一个合成的双产品报童问题和一个使用标普 500 资产回报的真实世界投资组合优化问题 。结果表明，对于有限支撑集问题，GTV 模糊集可以提供与 Wasserstein 和总变差模糊集等流行替代方案一样紧或更紧的泛化界 。因此，我们强调了将多种类型信息融入模糊集构建的实际好处，为不确定性下的数据驱动决策提供了改进的鲁棒性-性能权衡 。