Annals of Operations Research最新文献

Accurate forecasting of electricity consumption in residential buildings is essential for effective energy planning, real-time load balancing, and demand-side management. As residential demand becomes increasingly variable due to the proliferation of smart appliances and renewable energy systems, robust and adaptive forecasting models are crucial. To address this need, this study introduces a hybrid forecasting model that combines a deep Temporal Convolutional Network (TCN) with a Fuzzy Wavelet Neural Network (FWNN). The TCN component captures long-range temporal dependencies in time-series data, while the FWNN integrates fuzzy logic and wavelet transforms within neural network architecture, allowing the model to effectively handle uncertainty, imprecision, and nonlinearity commonly observed in electricity consumption behaviors. To further improve predictive accuracy, the model’s hyperparameters are fine-tuned using the Aquila Optimization metaheuristic algorithm. The proposed model is evaluated on two real-world datasets containing minute-level and hourly electricity consumption records. Comparative experiments with several established baseline models show that the suggested hybrid approach consistently outperforms its rivals across key performance metrics. These results underscore the model’s accuracy and robustness, positioning it as a promising candidate for integration into modern residential energy management systems.

Facility location problems on networks deal with locating facilities (e.g., parks, schools, or health facilities) to serve a set of clients (e.g., citizens). Many existing facility location studies assume that it is possible to (re-)locate predetermined or build new facilities. However, this assumption is not realistic for situations when relocating or building facilities would be too expensive or impossible. Recognizing this challenge, Berman et al. (Ann Oper Res 40:1–16, 1992) first propose the facility network addition modification problems (FNAMPs) on networks that aim to add a given number of new edges (e.g., constructing new roadways or bridges) to improve the client accessibility to the facilities by minimizing the total accessibility or maximum accessibility cost objectives based on clients’ distances to the facility. Yet, all existing approaches for FNAMPs are only heuristics, provide no solution quality guarantees, and fail to scale to even moderate-sized network instances. In this paper, we revisit a special case of FNAMPs with binary demands. We develop approximation algorithms and efficient heuristics for this special case under the two cost objectives. We then consider the strategic aspects of the special case of FNAMPs in which clients’ locations are private information. We design scalable strategyproof mechanisms to address FNAMPs under the two cost objectives while incentivizing clients to report their locations truthfully. We conduct extensive experiments on synthetic and real-world networks to demonstrate the effectiveness and efficiency of all proposed methods against various baselines as well as the strategyproofness of the proposed methods.

This paper introduces the special issue on Ensemble Learning for Operations Research and Business Analytics. Its main purpose is to provide summaries for the 14 contributing research papers that were accepted for inclusion in this special issue. We first define an updated and extended taxonomy of ensemble learner architectures to characterize and differentiate ensemble learning algorithms. Subsequently, we characterize the special issue contributions in two ways: with respect to the operations research (OR) application they address and contribute to, and methodologically with respect to the newly defined taxonomy. Finally, we present an ambitious agenda for future research on ensemble learning for OR and business analytics.

This paper reassesses the role of gold, silver, and palladium as hedging and safe-haven instruments against exchange rate risk, across both daily and intra-day frequencies. A novel, parsimonious testing framework is introduced to identify and model time-varying correlations between precious metals and currency returns, addressing common sources of overfitting and spurious dynamics in multivariate volatility modelling. Using data spanning key market stress episodes—such as the COVID-19 pandemic and the first phase of the Ukraine—the analysis examines the conditions under which precious metals provide effective hedges. The results highlight substantial heterogeneity across currencies and frequencies, with silver performing best at high frequencies and gold offering the most capital-efficient hedges for emerging markets during daily stress episodes. A newly added analysis quantifies the consequences of modelling spurious dynamic correlation, demonstrating that incorrect assumptions can significantly reduce hedging effectiveness. Findings yield stylized, empirically grounded recommendations for investors and policymakers seeking to manage currency risk using precious metals.

The article investigates the existence of a stable coalition structure in coalition partition games dealing with a system of distinct representatives. The system of distinct representatives, or the transversal, is a coalition of agents whose members represent different groups. An agent gains utility if he/she belongs to a transversal. The transversal value of a cooperative game with coalition structure is introduced to solve the coalition formation problem. It is demonstrated that any cooperative game has a partition that is simultaneously Nash stable, permutation stable, and total payoff maximization for the transversal value. The transversal value expresses the players’ payoff functions in the workgroup formation game and the game of chairpersons. A player’s payoff in the workgroup formation game is the monetary reward for the projects the player participated in. In the game of chairpersons, each player is interested in being the group’s leader. It is demonstrated that there exist stable coalition structures with a punctuality property in such games.

Recirculating aquaponics, the integration of aquaculture and hydroponics, presents a promising solution to global food security, resource scarcity, and environmental degradation. This paper examines a techno-economic analysis for large-scale commercial aquaponics investments, introducing a Bayesian Network simulation methodology designed to mitigate financial risks and improve investment efficiency, profitability, and sustainability of aquaponic operations. The methodological approach presented here effectively captures the intricacies of such large-scale investments. Furthermore, this paper demonstrates the application of scenario extraction from simulated data, enabling the comparison of outcomes under a variety of conditions. These conditions, which might include low and high market demand, the influence of certification price premiums, and various marketing strategies, can be flexibly combined to create multiple scenarios. The design of the simulation model allows for flexible generation and combination of these conditions, providing a wide range of potential scenarios for comparison and analysis. Indeed, this approach's flexibility and versatility make it an effective decision support system, particularly useful in complex domains characterized by many variables and high levels of uncertainty and risk. While the focus here is on aquaponics, the methodology can be applied to other domains, improving decision-making processes, and leading to more informed and robust conclusions.

A systematic survival analysis approach is developed to integrate the fundamental concepts from conventional statistics and the more explanatory, algorithmic, and computational techniques offered by optimization and machine learning. The proposed methodology enables the discovery of the combinations of significant features associated with time to an event of interest and produces combinatorial patterns in large-scale datasets. The combinatorial survival patterns are obtained as algorithmic solutions of a mixed integer linear programming problem. Application of the developed pattern-based survival analysis approach is demonstrated in benchmark medical datasets.

Long-term carpooling is a convenient and stable way for demanders who travel to their destinations for a long time and have similar travel time. How to match drivers and riders effectively is a very important problem in long-term carpooling. This paper proposes a stable matching method for long-term carpooling. Firstly, the stable matching problem of long-term carpooling is described, and the relevant definitions of stable matching are given. Secondly, a mixed-integer programming model is constructed with the objective of maximizing the total utility. Then, a heuristic algorithm based on knowledge rules and Benders decomposition is proposed. Finally, numerical experiments on different scales validate the feasibility and effectiveness of the proposed method. The results show that the price of stability is relatively small compared with system optimum. On this basis, we explore how certain parameters such as the stability constraints, objective function, cost-sharing method, vehicle capacity and maximum detour ratio, might affect the matching scheme.