Swarm and Evolutionary Computation最新文献_第9页

QACO: An adaptive multi-phase ant colony optimization approach for robust feature selection QACO：一种鲁棒特征选择的自适应多相蚁群优化方法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-19 DOI: 10.1016/j.swevo.2025.102217

Zihang Wang , Ye Liang , Wenwei Sun , Jiaming Liu

High-dimensional financial data contains numerous redundant or irrelevant indicators, which significantly reduce prediction accuracy and increase computational costs. Although ant colony optimization (ACO) has been widely adopted for wrapper feature selection, its traditional form often suffers from premature convergence when the search space is large and complex. Inspired by the ecological dynamics of real ant colonies, we introduce QACO, a multi-phase ACO framework that explicitly models three biological mechanisms: (i) dynamic colony resizing that reflects the changing queen/worker ratio, (ii) phase-based pheromone updating that follows the colony's life cycle, and (iii) adaptive offspring generation that gradually freezes high-impact features while preserving diversity. Extensive experiments on ten real-world Chinese corporate bankruptcy datasets and four tree-based learners, using a rigorous 30-run ten-fold cross-validation protocol, show that QACO achieves the highest average accuracy and the narrowest confidence intervals, outperforming eight state-of-the-art meta-heuristics in 88% of 360 statistical comparisons. Ablation analyses indicate that dynamic colony sizing speeds up early convergence by a factor of 3.4, while phase-based pheromone updating enhances recall by up to 4.8 percentage points without losing precision. The bio-inspired mechanisms proposed can be easily integrated into any swarm-based optimizer, offering a reliable and reusable solution for high-dimensional feature selection in financial risk prediction and beyond.

高维金融数据中包含大量冗余或不相关的指标，大大降低了预测精度，增加了计算成本。蚁群算法在包装器特征选择中得到了广泛的应用，但在搜索空间较大且复杂的情况下，其传统形式往往存在过早收敛的问题。受真实蚁群生态动态的启发，我们引入了QACO，这是一个多阶段蚁群算法框架，它明确地模拟了三种生物机制：(i)反映蚁后/工蚁比例变化的动态群体调整大小，（ii）遵循群体生命周期的基于阶段的信息素更新，以及（iii）在保持多样性的同时逐渐冻结高影响特征的适应性后代生成。在10个真实的中国企业破产数据集和4个基于树的学习器上进行的广泛实验，使用严格的30次交叉验证协议，表明QACO达到了最高的平均准确率和最窄的置信区间，在360次统计比较的88%中优于8个最先进的元启发式。消融分析表明，动态群体规模将早期收敛速度提高了3.4倍，而基于相位的信息素更新将召回率提高了4.8个百分点，同时不会失去精度。提出的仿生机制可以很容易地集成到任何基于群体的优化器中，为金融风险预测及其他领域的高维特征选择提供可靠和可重用的解决方案。

{"title":"QACO: An adaptive multi-phase ant colony optimization approach for robust feature selection","authors":"Zihang Wang , Ye Liang , Wenwei Sun , Jiaming Liu","doi":"10.1016/j.swevo.2025.102217","DOIUrl":"10.1016/j.swevo.2025.102217","url":null,"abstract":"<div><div>High-dimensional financial data contains numerous redundant or irrelevant indicators, which significantly reduce prediction accuracy and increase computational costs. Although ant colony optimization (ACO) has been widely adopted for wrapper feature selection, its traditional form often suffers from premature convergence when the search space is large and complex. Inspired by the ecological dynamics of real ant colonies, we introduce QACO, a multi-phase ACO framework that explicitly models three biological mechanisms: (i) dynamic colony resizing that reflects the changing queen/worker ratio, (ii) phase-based pheromone updating that follows the colony's life cycle, and (iii) adaptive offspring generation that gradually freezes high-impact features while preserving diversity. Extensive experiments on ten real-world Chinese corporate bankruptcy datasets and four tree-based learners, using a rigorous 30-run ten-fold cross-validation protocol, show that QACO achieves the highest average accuracy and the narrowest confidence intervals, outperforming eight state-of-the-art meta-heuristics in 88% of 360 statistical comparisons. Ablation analyses indicate that dynamic colony sizing speeds up early convergence by a factor of 3.4, while phase-based pheromone updating enhances recall by up to 4.8 percentage points without losing precision. The bio-inspired mechanisms proposed can be easily integrated into any swarm-based optimizer, offering a reliable and reusable solution for high-dimensional feature selection in financial risk prediction and beyond.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102217"},"PeriodicalIF":8.5,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571084","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Q-learning based memetic algorithm for flexible flow shop scheduling with active waiting and job switching 基于q学习的具有主动等待和作业切换的柔性流车间调度模因算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-19 DOI: 10.1016/j.swevo.2025.102227

Rui Xu , Yilin Xu , Wei Xiao

This paper investigates the flexible flow shop scheduling problem with active waiting and job switching, considering both inter-stage active waiting time and job switching cost. The study aims to optimize overall scheduling performance by simultaneously maximizing product quality, minimizing job switching cost, and maximizing delivery efficiency. To address the problem effectively, we formulate a mixed-integer programming model, and propose a Q-learning based memetic algorithm (MAQL). In the algorithm, a multi-stage dynamic decoding strategy is developed to handle the hybrid production scenarios involving both active-wait and nonactive-wait stages. Due to the NP-hardness of the problem, MAQL incorporates five local search operators, with Q-learning embedded to adaptively select the most appropriate one in each iteration to improve solution quality. In addition, an infeasible solution repair mechanism integrating three tailored strategies is designed to handle individuals that violate constraints during the search process. Comprehensive numerical experiments are conducted on 50 test instances in different scales, comparing MAQL with its variants and four other metaheuristic algorithms. The results show that MAQL improves solution quality by at least 3.7% in solution quality over the best-performing competitor, with the Friedman test confirming significant differences and MAQL achieving the top rank. Furthermore, a real-world case study from a tobacco enterprise is presented to validate the efficiency and effectiveness of MAQL.

研究了具有主动等待和作业切换的柔性流水车间调度问题，同时考虑了阶段间主动等待时间和作业切换成本。研究的目标是在产品质量最大化、工作转换成本最小化和交货效率最大化的同时，优化整体调度性能。为了有效地解决这个问题，我们建立了一个混合整数规划模型，并提出了一个基于q学习的模因算法（MAQL）。在该算法中，提出了一种多阶段动态解码策略，以处理包括活动-等待和非活动-等待阶段的混合生产场景。由于问题的np -硬度，MAQL结合了5个局部搜索算子，并嵌入Q-learning，在每次迭代中自适应选择最合适的算子，以提高解的质量。此外，设计了一种整合三种定制策略的不可行解修复机制来处理搜索过程中违反约束的个体。在50个不同尺度的测试实例上进行了综合数值实验，比较了MAQL及其变体和其他四种元启发式算法。结果表明，与表现最好的竞争对手相比，MAQL在溶液质量方面至少提高了3.7%，弗里德曼测试证实了显著差异，MAQL达到了最高水平。最后，以某烟草企业为例，验证了MAQL的效率和有效性。

{"title":"A Q-learning based memetic algorithm for flexible flow shop scheduling with active waiting and job switching","authors":"Rui Xu , Yilin Xu , Wei Xiao","doi":"10.1016/j.swevo.2025.102227","DOIUrl":"10.1016/j.swevo.2025.102227","url":null,"abstract":"<div><div>This paper investigates the flexible flow shop scheduling problem with active waiting and job switching, considering both inter-stage active waiting time and job switching cost. The study aims to optimize overall scheduling performance by simultaneously maximizing product quality, minimizing job switching cost, and maximizing delivery efficiency. To address the problem effectively, we formulate a mixed-integer programming model, and propose a Q-learning based memetic algorithm (MAQL). In the algorithm, a multi-stage dynamic decoding strategy is developed to handle the hybrid production scenarios involving both active-wait and nonactive-wait stages. Due to the NP-hardness of the problem, MAQL incorporates five local search operators, with Q-learning embedded to adaptively select the most appropriate one in each iteration to improve solution quality. In addition, an infeasible solution repair mechanism integrating three tailored strategies is designed to handle individuals that violate constraints during the search process. Comprehensive numerical experiments are conducted on 50 test instances in different scales, comparing MAQL with its variants and four other metaheuristic algorithms. The results show that MAQL improves solution quality by at least 3.7% in solution quality over the best-performing competitor, with the Friedman test confirming significant differences and MAQL achieving the top rank. Furthermore, a real-world case study from a tobacco enterprise is presented to validate the efficiency and effectiveness of MAQL.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102227"},"PeriodicalIF":8.5,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145536966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A multi-stage evolutionary algorithm based on finite state machine theory for constrained multi-objective optimization problem 基于有限状态机理论的约束多目标优化多阶段进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-19 DOI: 10.1016/j.swevo.2025.102222

Zhuoxuan Li, Zhaoguang Liu, Changgeng Su

Multi-stage optimization algorithms are promising for solving constrained multi-objective optimization problems (CMOPs), particularly when dealing with complex feasible regions. Although decomposing the problem into multiple stages does simplify the optimization process, the timing of stage switching is crucial. However, current approaches tend to focus on update strategies while lacking adequate emphasis on timely stage switching. This study introduces a multi-stage evolutionary algorithm based on finite state machine theory, named FSM-CMO. The algorithm employs a finite set of states with predictable switching rules, utilizing event/action-driven feedback to trigger timely shifts between customized update strategies. FSM-CMO operates through four states. In State I, the main population (MPop) uses the weight vectors of the Helping population (HPop) through restricted mating selection for initial exploration. In State II, HPop disregards constraints to drive MPop toward the unconstrained Pareto front. In State III, HPop employs a new sparsity evaluation strategy to assist MPop in gradually exploring the feasible boundary of the current region. Finally, in State IV, HPop uses the truncation strategy in SPEA2 to adjust the distribution of MPop. Extensive tests on four benchmark suites and fourteen real-world CMOPs have demonstrated its competitive performance compared to seven state-of-the-art algorithms. Its key strength lies in the state switching mechanism that flexibly schedules strategies based on population status, enhancing adaptability to complex constraints. Although the framework excels in tested scenarios, future work will target ultra-high-dimensional and dynamically constrained problems to enhance its practical applicability.

多阶段优化算法是求解约束多目标优化问题的有效方法，特别是在处理复杂可行区域时。虽然将问题分解为多个阶段确实简化了优化过程，但阶段切换的时机至关重要。然而，目前的方法往往侧重于更新策略，而对及时切换阶段缺乏足够的重视。提出了一种基于有限状态机理论的多阶段进化算法FSM-CMO。该算法采用有限的状态集和可预测的切换规则，利用事件/动作驱动的反馈来触发自定义更新策略之间的及时转换。FSM-CMO在四个州运营。在状态I中，主种群（MPop）通过限制性交配选择，利用辅助种群（HPop）的权重向量进行初始探索。在状态II中，HPop无视约束将MPop推向无约束的帕累托前沿。在状态III中，HPop采用一种新的稀疏度评价策略，帮助MPop逐步探索当前区域的可行边界。最后，在状态IV中，HPop使用SPEA2中的截断策略来调整MPop的分布。在四个基准套件和14个真实世界的cops上进行的广泛测试表明，与7种最先进的算法相比，它的性能具有竞争力。其关键优势在于状态切换机制，可以根据种群状态灵活调度策略，增强对复杂约束的适应性。尽管该框架在测试场景中表现出色，但未来的工作将针对超高维和动态约束问题，以增强其实际适用性。

{"title":"A multi-stage evolutionary algorithm based on finite state machine theory for constrained multi-objective optimization problem","authors":"Zhuoxuan Li, Zhaoguang Liu, Changgeng Su","doi":"10.1016/j.swevo.2025.102222","DOIUrl":"10.1016/j.swevo.2025.102222","url":null,"abstract":"<div><div>Multi-stage optimization algorithms are promising for solving constrained multi-objective optimization problems (CMOPs), particularly when dealing with complex feasible regions. Although decomposing the problem into multiple stages does simplify the optimization process, the timing of stage switching is crucial. However, current approaches tend to focus on update strategies while lacking adequate emphasis on timely stage switching. This study introduces a multi-stage evolutionary algorithm based on finite state machine theory, named FSM-CMO. The algorithm employs a finite set of states with predictable switching rules, utilizing event/action-driven feedback to trigger timely shifts between customized update strategies. FSM-CMO operates through four states. In State I, the main population (MPop) uses the weight vectors of the Helping population (HPop) through restricted mating selection for initial exploration. In State II, HPop disregards constraints to drive MPop toward the unconstrained Pareto front. In State III, HPop employs a new sparsity evaluation strategy to assist MPop in gradually exploring the feasible boundary of the current region. Finally, in State IV, HPop uses the truncation strategy in SPEA2 to adjust the distribution of MPop. Extensive tests on four benchmark suites and fourteen real-world CMOPs have demonstrated its competitive performance compared to seven state-of-the-art algorithms. Its key strength lies in the state switching mechanism that flexibly schedules strategies based on population status, enhancing adaptability to complex constraints. Although the framework excels in tested scenarios, future work will target ultra-high-dimensional and dynamically constrained problems to enhance its practical applicability.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102222"},"PeriodicalIF":8.5,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145536967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Decomposition-based multi-objective evolutionary algorithm for bi-optimal selection 基于分解的双最优选择多目标进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-18 DOI: 10.1016/j.swevo.2025.102228

Qinwei Fan , Dewei Yang , Jigen Peng , Haiyang Li , Jian Wang , Aoxue Yin

Decomposition-based multi-objective evolutionary algorithms often suffer from premature convergence when dealing with complex Pareto fronts. To address this issue, this paper proposes a decomposition-based bi-optional optimization algorithm (MOEA/D-BOS). The proposed method integrates the SPEA-II selection mechanism and an individual exploration strategy into the MOEA/D framework to enhance population diversity and information retention. In addition, a new weight vector generation method and a scalarization function are designed to improve the uniformity of solution distribution. Benchmark experiments demonstrate that, compared with several advanced algorithms, MOEA/D-BOS achieves superior performance in terms of both convergence speed and population diversity.

基于分解的多目标进化算法在处理复杂的Pareto前沿时往往存在过早收敛的问题。针对这一问题，本文提出了一种基于分解的双可选优化算法（MOEA/D-BOS）。该方法将SPEA-II选择机制和个体探索策略整合到MOEA/D框架中，以增强种群多样性和信息保留。此外，设计了一种新的权向量生成方法和标量化函数，提高了解分布的均匀性。基准实验表明，与几种先进的算法相比，MOEA/D-BOS在收敛速度和种群多样性方面都具有优越的性能。

引用次数: 0

Partition-informed Ant Colony Optimization for min–max multiple TSP 最小-最大多TSP的分区蚁群优化

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-18 DOI: 10.1016/j.swevo.2025.102224

Sara Pérez-Carabaza, Akemi Gálvez, Andrés Iglesias

The multiple traveling salesmen problem (mTSP) generalizes the classical TSP by involving multiple travelers who must collectively visit all cities, starting and ending at a common depot. This work focuses on the min–max variant, where the objective is to minimize the length of the longest subtour, ensuring a balanced workload among travelers, which is a crucial factor in many real-world applications, such as emergency response and logistics. This paper proposes a novel Ant Colony System (ACS)-based approach that effectively addresses the min–max mTSP, designed to construct well-balanced tours while optimizing the maximum tour length. The method integrates two key strategies: a sector-based heuristic for guiding city assignments, and a dynamic traveler selection criterion to promote equitable route construction. The method was evaluated on 33 two-dimensional Euclidean benchmark instances and compared with four state-of-the-art ACO-based approaches, demonstrating consistently better fitness under the min–max objective.

多旅行推销员问题（multiple traveling salesman problem, mTSP）是经典TSP问题的推广，它涉及多个旅行者，这些旅行者必须集体访问所有城市，从一个公共车站开始和结束。这项工作的重点是最小-最大变体，其目标是最小化最长子线路的长度，确保旅行者之间的工作量平衡，这是许多实际应用中的关键因素，例如应急响应和物流。本文提出了一种新的基于蚁群系统（ACS）的方法，该方法有效地解决了最小最大mTSP问题，旨在构建平衡的行程，同时优化最大行程长度。该方法集成了两个关键策略：基于扇区的启发式方法来指导城市分配，以及动态的旅行者选择标准来促进公平的路线建设。在33个二维欧几里得基准实例上对该方法进行了评估，并与四种最先进的基于aco的方法进行了比较，结果表明该方法在最小-最大目标下始终具有更好的适应度。

引用次数: 0

A sparse large-scale multi-objective evolutionary optimization based on bi-level interactive grouping 基于双层交互分组的稀疏大规模多目标进化优化

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-15 DOI: 10.1016/j.swevo.2025.102209

Yingjie Zou , Juan Zou , Shiting Wang , Yuan Liu , Shengxiang Yang

Sparse large-scale multi-objective optimization problems present a dual challenge: a vast number of decision variables and highly sparse Pareto-optimal sets, where traditional evolutionary algorithms often fail. To tackle these issues, we propose a Bi-level Interactive Grouping Evolutionary Algorithm (BLIGEA). The algorithm’s novelty lies in two main contributions. First, it introduces a bi-level interactive grouping strategy that applies distinct optimization mechanisms to the binary and real vectors of the solutions, fostering their synergistic co-evolution. Moreover, a knowledge-guided strategy is designed to effectively learn and leverage sparsity information from the population during the search process. Extensive experimental results on benchmarks and real-world applications demonstrate that BLIGEA significantly surpasses state-of-the-art methods.

稀疏大规模多目标优化问题提出了双重挑战：大量的决策变量和高度稀疏的帕累托最优集，这是传统进化算法经常失败的地方。为了解决这些问题，我们提出了一种双层交互分组进化算法（BLIGEA）。该算法的新颖性主要体现在两个方面。首先，它引入了一个双级交互分组策略，该策略将不同的优化机制应用于解决方案的二进制和实向量，促进它们的协同进化。此外，设计了一种知识引导策略，以便在搜索过程中有效地从种群中学习和利用稀疏性信息。在基准测试和实际应用中的广泛实验结果表明，BLIGEA显著优于最先进的方法。

引用次数: 0

A Dynamic Multi-objective Optimization approach for computing resource allocation in industrial model repository 工业模型库计算资源分配的动态多目标优化方法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-13 DOI: 10.1016/j.swevo.2025.102219

Dan Hu, Jianjun He

As Industry 4.0 and intelligent manufacturing continue to evolve, Industrial Model Repositories are playing an increasingly critical role in enabling model reuse, business collaboration, and intelligent decision-making. However, real-world implementations face substantial challenges due to unpredictable workload fluctuations, dynamic variations in resource availability and pricing, and complex inter-model dependencies. To address these issues, this paper formulates the scheduling task as a Dynamic Multi-objective Optimization Problem, proposing a resource scheduling framework that simultaneously minimizes consumption and maximizes utilization. The framework incorporates model subscription prediction and collaboration effect to accurately capture the temporal evolution and coupling features of model tasks. For the allocation algorithmic, a improved multi-objective evolutionary algorithm based on decomposition approach is developed, integrating diversity preservation, historical memory, and ARIMA-based prediction to enhance convergence and adaptability in dynamic environments. Experimental results across real-world experiment show that the proposed method significantly outperforms existing algorithms in terms of resource efficiency and solution stability. This research provides both theoretical foundations and practical strategies for intelligent resource management in industrial model platforms and offers promising insights for future optimization in complex industrial systems.

随着工业4.0和智能制造的不断发展，工业模型存储库在支持模型重用、业务协作和智能决策方面发挥着越来越重要的作用。然而，由于不可预测的工作负载波动、资源可用性和定价的动态变化以及复杂的模型间依赖关系，实际实现面临着巨大的挑战。为了解决这些问题，本文将调度任务表述为一个动态多目标优化问题，提出了一个消耗最小化和利用率最大化的资源调度框架。该框架结合模型订阅预测和协同效应，准确捕捉模型任务的时间演化和耦合特征。在分配算法方面，提出了一种改进的基于分解方法的多目标进化算法，将多样性保护、历史记忆和基于arima的预测相结合，增强了算法的收敛性和对动态环境的适应性。实际实验结果表明，该方法在资源效率和解的稳定性方面明显优于现有算法。本研究为工业模型平台的智能资源管理提供了理论基础和实践策略，为未来复杂工业系统的优化提供了有希望的见解。

{"title":"A Dynamic Multi-objective Optimization approach for computing resource allocation in industrial model repository","authors":"Dan Hu, Jianjun He","doi":"10.1016/j.swevo.2025.102219","DOIUrl":"10.1016/j.swevo.2025.102219","url":null,"abstract":"<div><div>As Industry 4.0 and intelligent manufacturing continue to evolve, Industrial Model Repositories are playing an increasingly critical role in enabling model reuse, business collaboration, and intelligent decision-making. However, real-world implementations face substantial challenges due to unpredictable workload fluctuations, dynamic variations in resource availability and pricing, and complex inter-model dependencies. To address these issues, this paper formulates the scheduling task as a Dynamic Multi-objective Optimization Problem, proposing a resource scheduling framework that simultaneously minimizes consumption and maximizes utilization. The framework incorporates model subscription prediction and collaboration effect to accurately capture the temporal evolution and coupling features of model tasks. For the allocation algorithmic, a improved multi-objective evolutionary algorithm based on decomposition approach is developed, integrating diversity preservation, historical memory, and ARIMA-based prediction to enhance convergence and adaptability in dynamic environments. Experimental results across real-world experiment show that the proposed method significantly outperforms existing algorithms in terms of resource efficiency and solution stability. This research provides both theoretical foundations and practical strategies for intelligent resource management in industrial model platforms and offers promising insights for future optimization in complex industrial systems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"99 ","pages":"Article 102219"},"PeriodicalIF":8.5,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145519889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Simulation-based multi-objective optimization for mission specific tuning of swarming robots 基于仿真的蜂群机器人多目标任务优化

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-11 DOI: 10.1016/j.swevo.2025.102215

Reda Ghanem , Ismail M. Ali , Kathryn Kasmarik , Matthew Garratt

Swarm robotics has been applied to various applications, including exploration, task allocation, and coverage problems. This application is challenging because decision-makers must calibrate swarming collective motion parameters while maintaining priority for mission-specific objectives. This paper presents a novel simulation-based multi-objective optimization framework that autonomously tunes collective motion parameters for swarming robots solving coverage problems. Our proposed Nondominated Sorting Genetic Algorithm II with Mixed Crossover Mutation (NSGA2-MCM) approach permits decision-makers to balance competing objectives by selecting optimal swarm parameters for specific mission requirements. We evaluate our approach against state-of-the-art multi-objective optimization methods using established performance metrics. Results show that our algorithm boosts Hypervolume by up to 10.82% and cuts Generational Distance by up to 7.62%. As environment complexity increases, it achieves Hypervolume gains of up to 49.98% and Generational Distance reductions of up to 24.28%. Furthermore, C-metric analysis reveals that NSGA2-MCM dominates an average of 88.89% of alternative algorithms’ solutions.

蜂群机器人已经应用于各种应用，包括探索、任务分配和覆盖问题。这种应用具有挑战性，因为决策者必须校准群体集体运动参数，同时保持任务特定目标的优先级。提出了一种新的基于仿真的多目标优化框架，该框架可自主调整群体机器人的集体运动参数，用于解决覆盖问题。我们提出的非支配排序遗传算法II混合交叉突变（NSGA2-MCM）方法允许决策者通过选择特定任务要求的最优群体参数来平衡竞争目标。我们使用既定的性能指标来评估我们的方法，以对抗最先进的多目标优化方法。结果表明，该算法将Hypervolume提升了10.82%，将代际距离降低了7.62%。随着环境复杂性的增加，它实现了高达49.98%的Hypervolume增益和高达24.28%的代际距离减少。此外，C-metric分析显示，NSGA2-MCM平均占替代算法解决方案的88.89%。

{"title":"Simulation-based multi-objective optimization for mission specific tuning of swarming robots","authors":"Reda Ghanem , Ismail M. Ali , Kathryn Kasmarik , Matthew Garratt","doi":"10.1016/j.swevo.2025.102215","DOIUrl":"10.1016/j.swevo.2025.102215","url":null,"abstract":"<div><div>Swarm robotics has been applied to various applications, including exploration, task allocation, and coverage problems. This application is challenging because decision-makers must calibrate swarming collective motion parameters while maintaining priority for mission-specific objectives. This paper presents a novel simulation-based multi-objective optimization framework that autonomously tunes collective motion parameters for swarming robots solving coverage problems. Our proposed Nondominated Sorting Genetic Algorithm II with Mixed Crossover Mutation (NSGA2-MCM) approach permits decision-makers to balance competing objectives by selecting optimal swarm parameters for specific mission requirements. We evaluate our approach against state-of-the-art multi-objective optimization methods using established performance metrics. Results show that our algorithm boosts Hypervolume by up to 10.82% and cuts Generational Distance by up to 7.62%. As environment complexity increases, it achieves Hypervolume gains of up to 49.98% and Generational Distance reductions of up to 24.28%. Furthermore, C-metric analysis reveals that NSGA2-MCM dominates an average of 88.89% of alternative algorithms’ solutions.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"99 ","pages":"Article 102215"},"PeriodicalIF":8.5,"publicationDate":"2025-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145519885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-objective evolutionary feature selection for ensemble learning with random forests in time series forecasting 时间序列预测中随机森林集成学习多目标进化特征选择

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-10 DOI: 10.1016/j.swevo.2025.102211

Raquel Espinosa , Gracia Sánchez , José Palma , Fernando Jiménez

Time series forecasting is fundamental in numerous domains, including finance, healthcare, energy, and environmental monitoring. However, the high dimensionality of feature spaces can lead to overfitting and reduced interpretability, making feature selection a crucial preprocessing step. This paper proposes a multi-objective evolutionary algorithm for feature selection in time series forecasting, designed to enhance predictive accuracy while improving generalization. The method partitions the dataset, associating each partition with an objective function in the optimization process. By independently selecting relevant feature subsets, it generates a Pareto front of Random Forest models, each trained on a distinct subset of features. These models are then aggregated into a stacking-based ensemble framework, effectively balancing feature relevance and diversity. Additionally, we introduce a feature importance measure based on selection frequency in the non-dominated solutions of the optimization process. To validate our approach, we conduct experiments on real-world forecasting tasks, including air quality prediction in southeastern Spain and Italy and oil temperature forecasting in industrial applications. We also evaluate performance on synthetic datasets of increasing complexity, systematically varying instances, features, seasonality, noise, and trends. The proposed method is compared against conventional Random Forest, a wrapper-based feature selection method with a multi-objective evolutionary search strategy, and several state-of-the-art embedded feature selection techniques for time series forecasting. The results demonstrate that our approach significantly improves forecasting accuracy while mitigating overfitting. By integrating multi-objetive evolutionary optimization, random forest, ensemble learning, and a novel feature importance measure, our method offers a robust, interpretable, and effective feature selection for time series forecasting applications.

时间序列预测是许多领域的基础，包括金融、医疗保健、能源和环境监测。然而，特征空间的高维可能导致过拟合和可解释性降低，使得特征选择成为关键的预处理步骤。本文提出了一种多目标进化算法用于时间序列预测的特征选择，旨在提高预测精度的同时提高泛化能力。该方法对数据集进行分区，并在优化过程中将每个分区与目标函数相关联。通过独立选择相关的特征子集，生成随机森林模型的帕累托前，每个模型都在一个不同的特征子集上训练。然后将这些模型聚合到基于堆栈的集成框架中，有效地平衡特征相关性和多样性。此外，我们在优化过程的非支配解中引入了基于选择频率的特征重要性度量。为了验证我们的方法，我们对现实世界的预测任务进行了实验，包括西班牙东南部和意大利的空气质量预测以及工业应用中的油温预测。我们还评估了越来越复杂的合成数据集的性能，系统地改变实例、特征、季节性、噪声和趋势。将该方法与传统随机森林、一种基于包装的多目标进化搜索策略特征选择方法以及几种最新的用于时间序列预测的嵌入式特征选择技术进行了比较。结果表明，我们的方法在减少过拟合的同时显著提高了预测精度。通过集成多目标进化优化、随机森林、集成学习和一种新的特征重要性度量，我们的方法为时间序列预测应用提供了一个鲁棒的、可解释的和有效的特征选择。

{"title":"Multi-objective evolutionary feature selection for ensemble learning with random forests in time series forecasting","authors":"Raquel Espinosa , Gracia Sánchez , José Palma , Fernando Jiménez","doi":"10.1016/j.swevo.2025.102211","DOIUrl":"10.1016/j.swevo.2025.102211","url":null,"abstract":"<div><div>Time series forecasting is fundamental in numerous domains, including finance, healthcare, energy, and environmental monitoring. However, the high dimensionality of feature spaces can lead to overfitting and reduced interpretability, making feature selection a crucial preprocessing step. This paper proposes a multi-objective evolutionary algorithm for feature selection in time series forecasting, designed to enhance predictive accuracy while improving generalization. The method partitions the dataset, associating each partition with an objective function in the optimization process. By independently selecting relevant feature subsets, it generates a Pareto front of Random Forest models, each trained on a distinct subset of features. These models are then aggregated into a stacking-based ensemble framework, effectively balancing feature relevance and diversity. Additionally, we introduce a feature importance measure based on selection frequency in the non-dominated solutions of the optimization process. To validate our approach, we conduct experiments on real-world forecasting tasks, including air quality prediction in southeastern Spain and Italy and oil temperature forecasting in industrial applications. We also evaluate performance on synthetic datasets of increasing complexity, systematically varying instances, features, seasonality, noise, and trends. The proposed method is compared against conventional Random Forest, a wrapper-based feature selection method with a multi-objective evolutionary search strategy, and several state-of-the-art embedded feature selection techniques for time series forecasting. The results demonstrate that our approach significantly improves forecasting accuracy while mitigating overfitting. By integrating multi-objetive evolutionary optimization, random forest, ensemble learning, and a novel feature importance measure, our method offers a robust, interpretable, and effective feature selection for time series forecasting applications.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"99 ","pages":"Article 102211"},"PeriodicalIF":8.5,"publicationDate":"2025-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145519887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A consensus-based EDA with multi-scale neighborhood search for vehicle routing problem with pickup and delivery 基于共识的多尺度邻域搜索的车辆取货路径问题EDA

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2025-11-10 DOI: 10.1016/j.swevo.2025.102216

Wei Wang , Yindong Shen

The vehicle routing problem with pickup and delivery (VRPPD) is concerned with planning optimal routes for a fleet of vehicles to meet the diverse demands of customers. In this problem, scenarios involving either simultaneous pickup and delivery (SPD) or mixed pickup and delivery (MPD) cause fluctuations in vehicle loads. Our theoretical analyses reveal that SPD tightens vehicle capacity constraints, reducing the number of feasible solutions, while MPD expands the feasible region, thus increasing the number of local optima. This necessitates a generic algorithm to balance diversification and intensification and to promote persistent exploration. Therefore, this paper considers multiple neighborhood scales and proposes a consensus-based estimation of distribution algorithm (EDA) incorporating the scalable large neighborhood search (SLNS) and the tour fragment recombination (TFR), abbreviated as CEDA-ST. In the CEDA-ST, population consensus is leveraged to estimate the distribution of optimal routes, generating a consensus matrix for individual construction and neighborhood searches. The SLNS operator conducts destroy-and-repair moves in large neighborhoods to promote diversification. Meanwhile, the TFR operator facilitates local improvements in small neighborhoods to enhance intensification. Furthermore, a stagnation-triggered diversity management (STDM) strategy is developed to eliminate redundant individuals, encouraging persistent exploration. Comparative experiments demonstrate its superiority. An effectiveness analysis and two ablation experiments highlight the contributions of consensus information and multi-scale neighborhood search, respectively. Additionally, a real-world case study on JD Logistics further validates the applicability of CEDA-ST in practical scenarios.

取货车辆路线问题（VRPPD）关注的是为车队规划最优路线，以满足客户的不同需求。在这个问题中，涉及同时拾取和交付（SPD）或混合拾取和交付（MPD）的场景会导致车辆负载的波动。理论分析表明，SPD收紧了车辆容量约束，减少了可行解的数量；MPD扩大了可行区域，增加了局部最优解的数量。这就需要一种通用算法来平衡多样化和集约化，并促进持续的探索。因此，本文考虑多个邻域尺度，提出了一种基于共识的分布估计算法（EDA），该算法结合了可扩展大邻域搜索（SLNS）和游片段重组（TFR），简称CEDA-ST。在cda - st中，利用人口共识来估计最优路线的分布，生成单个建筑和邻居搜索的共识矩阵。SLNS运营商在大型社区进行摧毁和修复行动，以促进多样化。同时，TFR运营商促进小社区的局部改善，以加强集约化。此外，我们还开发了一种停滞触发多样性管理（STDM）策略，以消除冗余的个体，鼓励持续的探索。对比实验证明了其优越性。有效性分析和两个消融实验分别突出了共识信息和多尺度邻域搜索的贡献。此外，通过对京东物流的实际案例研究，进一步验证了cda - st在实际场景中的适用性。

{"title":"A consensus-based EDA with multi-scale neighborhood search for vehicle routing problem with pickup and delivery","authors":"Wei Wang , Yindong Shen","doi":"10.1016/j.swevo.2025.102216","DOIUrl":"10.1016/j.swevo.2025.102216","url":null,"abstract":"<div><div>The vehicle routing problem with pickup and delivery (VRPPD) is concerned with planning optimal routes for a fleet of vehicles to meet the diverse demands of customers. In this problem, scenarios involving either simultaneous pickup and delivery (SPD) or mixed pickup and delivery (MPD) cause fluctuations in vehicle loads. Our theoretical analyses reveal that SPD tightens vehicle capacity constraints, reducing the number of feasible solutions, while MPD expands the feasible region, thus increasing the number of local optima. This necessitates a generic algorithm to balance diversification and intensification and to promote persistent exploration. Therefore, this paper considers multiple neighborhood scales and proposes a consensus-based estimation of distribution algorithm (EDA) incorporating the scalable large neighborhood search (SLNS) and the tour fragment recombination (TFR), abbreviated as CEDA-ST. In the CEDA-ST, population consensus is leveraged to estimate the distribution of optimal routes, generating a consensus matrix for individual construction and neighborhood searches. The SLNS operator conducts destroy-and-repair moves in large neighborhoods to promote diversification. Meanwhile, the TFR operator facilitates local improvements in small neighborhoods to enhance intensification. Furthermore, a stagnation-triggered diversity management (STDM) strategy is developed to eliminate redundant individuals, encouraging persistent exploration. Comparative experiments demonstrate its superiority. An effectiveness analysis and two ablation experiments highlight the contributions of consensus information and multi-scale neighborhood search, respectively. Additionally, a real-world case study on JD Logistics further validates the applicability of CEDA-ST in practical scenarios.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"99 ","pages":"Article 102216"},"PeriodicalIF":8.5,"publicationDate":"2025-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145519886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0