Annals of Operations Research最新文献

The integration of Artificial Intelligence (AI) within financial institutions has accelerated in response to mounting complexities and recurrent global crises, revealing the well-known but fair limitations of traditional economic models. Enhanced computational capacity and Big Data (BD) availability have enabled the adoption of Machine Learning (ML) across fields from banking, asset and risk management, and insurance, for tasks such as credit scoring, fraud detection, and market surveillance. AI facilitates portfolio construction and risk budgeting by synthesising historical and alternative data, while Natural Language Processing (NLP) aids in interpreting regulatory texts and central bank communications. However, the opacity of AI models introduces novel risks, including among others validation challenges and systemic vulnerabilities due to model convergence. When controlled appropriately, AI can paradoxically serve both as a risk source and a mitigation tool. This special issue presents rigorously peer-reviewed contributions that explore AI, ML, and BD applications in asset pricing, risk management, macroeconomic forecasting, and sustainable finance, offering valuable insights for academics, practitioners, and policymakers navigating financial uncertainty.

This special issue of Annals of Operations Research presents a collection of 31 papers that emerged from the 18th European Meeting on Game Theory (SING 18), held in Messina, Italy, in June 2023. The contributions span the full breadth of game-theoretical research, from foundational developments in cooperative and non-cooperative game theory to innovative applications in supply chain management, healthcare, finance, environmental policy, and emerging technologies. This collection demonstrates the vitality and relevance of game theory in addressing contemporary challenges across diverse domains, highlighting the synergy between theoretical advances and practical problem-solving. The papers showcase the evolution of game-theoretical methods, incorporating modern computational techniques including artificial intelligence, deep learning, and quantum computing, while maintaining strong connections to classical game-theoretic principles.

This research investigates the bounded batch online-scheduling issue, specifically focusing on incompatible job families assigned to unit flowshop machines. The main criterion is to reduce the makespan. Within a unit flowshop setting, each machine standardizes the processing time of a job to one unit. The concept of linear lookahead pertains to an online algorithm’s ability to anticipate job information within the time interval ((t, lambda t +beta ]) at time t. The bound batch capacity, denoted by (b < infty), signifies a restriction on the number of jobs that can be contained within a batch. For two unit machines, we introduce the optimal online algorithm (A^{b}). In instances where multiple unit machines are employed with parameters (lambda ge 1), (0 le beta < 1), and (f ge 2), we establish a competitive ratio bounded by at most (1 + max left{ {frac{f}{{(2f - 1)(1 + alpha )}},frac{f}{{lambda (2f - 1)alpha + beta + f}}} right}) based on the online algorithm (A_{f}^{b}). Specifically, we present an optimal online algorithm when (lambda =1) and (0 le beta le displaystyle frac{3f + 1-sqrt{9 f^{2}-2f +1}}{4}). Additionally, we extend our findings to the bounded batch scheduling problem incorporating a linear lookahead interval and (f (=m)) incompatible job families on multiple unit machines.

A model of state-contingent multi-input-output technologies, with a netput representation, is introduced to analyze different types of efficiencies. Overall efficiency allocative efficiency and technical efficiency are studied according to coalitions of firms in a firm game (being a cooperative TU-game). Different risks are investigated, inherent to either inputs or outputs, in order to measure their impacts on the different efficiencies. The risks imposed on the random direction of the directional distance function is axiomatically characterized to provide an operational functional form to measure efficiency under risks. An empirical application in portfolio management is presented, focusing on the performance evaluation of portfolios with different risk parameterization of the random direction.

In the context of global climate change and resource constraints, the sustainability of supply chain management has attracted widespread attention across various sectors. This special issue highlights the cutting-edge applications of operations research methods in sustainable supply chain management, particularly within the contexts of digitalization, low-carbon initiatives, and the circular economy. It showcases advances in supply chain resilience modeling enabled by blockchain and machine learning, low-carbon closed-loop game theory, green investment and subsidy strategies, AI-driven low-carbon transformation of small and medium-sized enterprises, and risk mitigation and performance evaluation of circular supply chains. These studies not only extend the theoretical boundaries of supply chain management amid digitalization and the circular economy but also provide practitioners with actionable decision-making tools. Future research should emphasize the integration of interdisciplinary approaches, multi-agent collaborative modeling, and the combination of AI with system dynamics to foster balanced and sustainable development of supply chains across environmental, economic, and social dimensions. Consequently, this special issue establishes a solid foundation for operations research addressing climate change and efficient resource utilization in supply chains.

We introduce the MinCovTarget+ algorithm for the problem of fair allocation of indivisible items and we study its performance with respect to some popular fairness and efficiency criteria such as minimal envy, proportionality, maximal Nash welfare and maximal total welfare. By a detailed numerical analysis we compare our newly proposed algorithm with a standard algorithm for this kind of problem: the Spliddit algorithm. Our numerical analysis shows that MinCovTarget+ provides allocations with an excellent balance between fairness and efficiency criteria. In particular, it typically yields minimal (null) envy solutions with a very high value of Nash welfare, at a fraction of the computation time used by Spliddit. Moreover, MinCovTarget+ can be applied in higher dimensions where Spliddit cannot be readily implemented. Our paper is the first in the literature to present a numerical study of these algorithms using a random uniform valuation of the goods to be allocated, as well as a novel design of the value matrix that incorporates dependent valuations. All the numerical estimates in this paper were obtained using a Macbook Air (Apple M1, 8 GB RAM). The corresponding MATLAB code is available at: https://github.com/giovannipuccetti/MinCovTarget. A user-friendly version of the algorithm is available at https://www.fair-allocation.com.

Pricing a global product differently across multiple regions is a common but controversial practice. Although price differentiation helps capture unique market characteristics, it also encourages parallel trade, which may affect the overall corporate performance of a global company. We study this problem with a single global business unit (GBU) and multiple local business units (LBUs). The GBU manufactures a product and sets a transfer price for supplying the product to all LBUs, and LBUs decide retail prices for their respective regional markets. Customers can purchase products in any region by comparing LBUs’ prices and other parallel-imported factors, and we construct the demand by a mixed-multinomial logit model. Therefore, each LBU needs to consider the other LBUs’ prices when setting its price, and we formulate this problem as a two-stage game-theoretic model. We verify the existence of a pure-strategy Nash equilibrium. We then develop a learning-based algorithm to find the equilibrium prices of LBUs. Our algorithm is computationally efficient with a large scale of decision-makers. Even for cases with 100 decision-makers, the pure-strategy Nash equilibrium can be obtained within 30 minutes. Numerical studies are conducted using data from the fast-moving consumer products industries. Although parallel trade is detrimental to some LBUs, the existence of parallel trade surprisingly leads to higher overall corporate profits, which is made possible by making the product available at different price points in a market. The proposed method also enables the quantitative validation of several conjectures on parallel trading practices.