Expert Systems with Applications最新文献

A reinforcement learning-driven hyper-heuristic algorithm for haul transportation and terminal delivery optimization in two-echelon distribution systems: A case study in GBA 基于强化学习驱动的两级配送系统运输和终端配送优化超启发式算法：以大湾区为例

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-16 DOI: 10.1016/j.eswa.2026.131206

Zhi Tang , Ting Qu , Yongheng Zhang , Yanghua Pan , Liqiang Ding , George Q. Huang

The two-echelon distribution system has been increasingly adopted in modern e-commerce logistics. However, as customer service requirements become more diverse, logistics providers must simultaneously accommodate delivery and pickup requests while also satisfying multiple non-overlapping time windows. These additional constraints substantially increase the complexity of balancing service quality and operational costs. To address this challenge, this study extends the classical two-echelon vehicle routing problem and introduces a new variant, called the two-echelon vehicle routing problem with simultaneous pickup and delivery under multiple time windows (2E-VRPSPDMTW). A mixed-integer linear programming (MILP) model is formulated to minimize total operational cost. Given the NP-hard nature of the problem, a Q-learning-based hyper-heuristic algorithm (QLHHA) is developed. The proposed framework first applies a spatiotemporal clustering strategy to allocate customers to satellites, thereby reducing the search space. It then constructs a pool of eight low-level heuristic operators, while a Q-learning mechanism serves as the high-level controller to adaptively select the most appropriate operator. A case study based on real operational data from a cross-border e-commerce logistics company in the Guangdong-Hong Kong-Macau Greater Bay Area (GBA) is conducted to evaluate the method. Extensive test cases and ablation experiment results demonstrate that QLHHA surpasses several state-of-the-art algorithms in both solution quality and stability, achieving up to a 10% reduction in total operational cost. Sensitivity analyses further reveal that, for large-scale demand scenarios, moderately widening the time-window width can substantially reduce operational cost.

现代电子商务物流越来越多地采用两级配送体系。然而，随着客户服务需求变得越来越多样化，物流供应商必须同时满足送货和取件请求，同时还要满足多个不重叠的时间窗口。这些额外的限制大大增加了平衡服务质量和运营成本的复杂性。为了解决这一挑战，本研究扩展了经典的两梯队车辆路线问题，并引入了一个新的变体，称为多时间窗口下同时取货和交货的两梯队车辆路线问题（2E-VRPSPDMTW）。建立了一个混合整数线性规划（MILP）模型，以最小化总运行成本。考虑到问题的NP-hard性质，开发了一种基于q学习的超启发式算法（QLHHA）。该框架首先采用时空聚类策略将客户分配给卫星，从而减少了搜索空间。然后，它构建了一个由8个低级启发式算子组成的池，而q -学习机制作为高级控制器自适应选择最合适的算子。以粤港澳大湾区（GBA）跨境电商物流公司的实际运营数据为例，对该方法进行了评估。大量的测试案例和消融实验结果表明，QLHHA在溶液质量和稳定性方面都超过了几种最先进的算法，可将总运营成本降低10%。敏感性分析进一步表明，对于大规模需求情景，适度扩大时间窗宽度可以大幅降低运营成本。

{"title":"A reinforcement learning-driven hyper-heuristic algorithm for haul transportation and terminal delivery optimization in two-echelon distribution systems: A case study in GBA","authors":"Zhi Tang , Ting Qu , Yongheng Zhang , Yanghua Pan , Liqiang Ding , George Q. Huang","doi":"10.1016/j.eswa.2026.131206","DOIUrl":"10.1016/j.eswa.2026.131206","url":null,"abstract":"<div><div>The two-echelon distribution system has been increasingly adopted in modern e-commerce logistics. However, as customer service requirements become more diverse, logistics providers must simultaneously accommodate delivery and pickup requests while also satisfying multiple non-overlapping time windows. These additional constraints substantially increase the complexity of balancing service quality and operational costs. To address this challenge, this study extends the classical two-echelon vehicle routing problem and introduces a new variant, called the two-echelon vehicle routing problem with simultaneous pickup and delivery under multiple time windows (2E-VRPSPDMTW). A mixed-integer linear programming (MILP) model is formulated to minimize total operational cost. Given the NP-hard nature of the problem, a Q-learning-based hyper-heuristic algorithm (QLHHA) is developed. The proposed framework first applies a spatiotemporal clustering strategy to allocate customers to satellites, thereby reducing the search space. It then constructs a pool of eight low-level heuristic operators, while a Q-learning mechanism serves as the high-level controller to adaptively select the most appropriate operator. A case study based on real operational data from a cross-border e-commerce logistics company in the Guangdong-Hong Kong-Macau Greater Bay Area (GBA) is conducted to evaluate the method. Extensive test cases and ablation experiment results demonstrate that QLHHA surpasses several state-of-the-art algorithms in both solution quality and stability, achieving up to a 10% reduction in total operational cost. Sensitivity analyses further reveal that, for large-scale demand scenarios, moderately widening the time-window width can substantially reduce operational cost.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"309 ","pages":"Article 131206"},"PeriodicalIF":7.5,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145981725","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

Intelligent neural architecture search via Taguchi design and language model-based differential evolution for agricultural image recognition 基于田口设计的智能神经架构搜索和基于语言模型的差异进化农业图像识别

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-16 DOI: 10.1016/j.eswa.2026.131135

Debtanu Ghosh , Subhayu Ghosh , Nanda Dulal Jana , Rammohan Mallipeddi

Vision Transformers (ViTs) have gained significant attention for their ability to achieve state-of-the-art performance in various computer vision tasks. However, designing an optimal ViT architecture requires extensive domain expertise and computational resources, making it more challenging for researchers. The selection of hyperparameters plays a critical role in model performance, yet determining the most influential parameters remains a complex task. To address this, we propose a two-stage Neural Architecture Search (NAS) framework that integrates Taguchi method based Design of Experiments (DoE) and Differential Evolution (DE) enhanced by Large Language Model (LLM)-based crossover. The DoE-Taguchi method is employed to systematically analyze the importance of hyperparameters and rank them based on their influence on model performance. The ranked parameters are then used to guide the DE, where an LLM-assisted crossover mechanism enhances exploration and convergence, leading to the discovery of near optimal architectures. The proposed approach is trained and tested on three different agricultural images datasets. The experimental results demonstrate the effectiveness of DoE and DE-LLM based optimized ViT models by showcasing their ability to handle the unique complexities of agricultural datasets while achieving superior accuracy and reliability in classification tasks.

视觉变压器（ViTs）因其在各种计算机视觉任务中实现最先进性能的能力而获得了极大的关注。然而，设计最佳的ViT架构需要广泛的领域专业知识和计算资源，这对研究人员来说更具挑战性。超参数的选择在模型性能中起着至关重要的作用，但确定最具影响力的参数仍然是一项复杂的任务。为了解决这个问题，我们提出了一个两阶段神经结构搜索（NAS）框架，该框架集成了基于田口方法的实验设计（DoE）和基于大语言模型（LLM）的交叉增强的差分进化（DE）。采用DoE-Taguchi方法系统地分析了超参数的重要性，并根据它们对模型性能的影响对它们进行了排序。然后使用排名参数来指导DE，其中llm辅助的交叉机制增强了探索和收敛，从而发现接近最优的架构。该方法在三个不同的农业图像数据集上进行了训练和测试。实验结果证明了基于DoE和DE-LLM的优化ViT模型的有效性，展示了它们处理农业数据集独特复杂性的能力，同时在分类任务中实现了卓越的准确性和可靠性。

{"title":"Intelligent neural architecture search via Taguchi design and language model-based differential evolution for agricultural image recognition","authors":"Debtanu Ghosh , Subhayu Ghosh , Nanda Dulal Jana , Rammohan Mallipeddi","doi":"10.1016/j.eswa.2026.131135","DOIUrl":"10.1016/j.eswa.2026.131135","url":null,"abstract":"<div><div>Vision Transformers (ViTs) have gained significant attention for their ability to achieve state-of-the-art performance in various computer vision tasks. However, designing an optimal ViT architecture requires extensive domain expertise and computational resources, making it more challenging for researchers. The selection of hyperparameters plays a critical role in model performance, yet determining the most influential parameters remains a complex task. To address this, we propose a two-stage Neural Architecture Search (NAS) framework that integrates Taguchi method based Design of Experiments (DoE) and Differential Evolution (DE) enhanced by Large Language Model (LLM)-based crossover. The DoE-Taguchi method is employed to systematically analyze the importance of hyperparameters and rank them based on their influence on model performance. The ranked parameters are then used to guide the DE, where an LLM-assisted crossover mechanism enhances exploration and convergence, leading to the discovery of near optimal architectures. The proposed approach is trained and tested on three different agricultural images datasets. The experimental results demonstrate the effectiveness of DoE and DE-LLM based optimized ViT models by showcasing their ability to handle the unique complexities of agricultural datasets while achieving superior accuracy and reliability in classification tasks.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"309 ","pages":"Article 131135"},"PeriodicalIF":7.5,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145981843","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

Human-AI collaborative scoring strategy of subjective assignments considering learning and fatigue effects 考虑学习和疲劳效应的主观作业人机协同评分策略

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-15 DOI: 10.1016/j.eswa.2026.131168

Qian Wang , Yan Wan , Feng Feng , Xiaokang Wang , Zhiyan Zhang

Although existing studies have focused on efficiency promotion or threshold optimization in human-AI collaborative scoring for subjective assignments, there has been relatively little attention given to how changes in human capabilities and psychological states affect system performance. To fill this research gap, we present a novel simulation framework for human-AI collaborative scoring strategy from a view point of cognitive ergonomics. Firstly, we design a rater ability model with an associated score mapping function to simulate human performance under learning and fatigue effects. Secondly, we develop a three-objective optimization model to balance human costs, scoring fairness, and scoring consistency under the scenarios of learning and fatigue effects. Finally, considering decision-makers’ loss aversion psychology from behavioral economics, we utilize the behavioral technique for order of preference by similarity to ideal solution (TOPSIS) approach to rank human-AI collaboration schemes. In addition, performance comparisons and parametric analyses on eight prompts from the automated student assessment prize (ASAP) data set are conducted to demonstrate the feasibility and superiority of the proposed framework. More specifically, under the learning effect, an expert proportion of 6.39% leads to an average fairness gain of 62.93%, whereas under the fatigue effect, an expert proportion of 4.12% results in an average fairness gain of 61.07%. It is important to note that this work offers a conceptual simulation framework whose practical deployment necessitates empirical validation of its behavioral assumptions.

虽然现有的研究主要集中在人类-人工智能主观作业协同评分的效率提升或阈值优化上，但对人类能力和心理状态的变化如何影响系统性能的关注相对较少。为了填补这一研究空白，我们从认知工效学的角度提出了一种新的人类-人工智能协同计分策略仿真框架。首先，我们设计了一个带有评分映射函数的评分能力模型来模拟人类在学习和疲劳作用下的表现。其次，在学习效应和疲劳效应两种情况下，建立了人力成本、评分公平性和评分一致性的三目标优化模型。最后，从行为经济学角度考虑决策者的损失厌恶心理，利用TOPSIS （similarity of preference order of ideal solution）方法对人机协作方案进行排序。此外，对来自自动学生评估奖（ASAP）数据集的八个提示进行了性能比较和参数分析，以证明所提出框架的可行性和优越性。更具体地说，在学习效应下，专家比例为6.39%，平均公平收益为62.93%，而在疲劳效应下，专家比例为4.12%，平均公平收益为61.07%。值得注意的是，这项工作提供了一个概念模拟框架，其实际部署需要对其行为假设进行经验验证。

{"title":"Human-AI collaborative scoring strategy of subjective assignments considering learning and fatigue effects","authors":"Qian Wang , Yan Wan , Feng Feng , Xiaokang Wang , Zhiyan Zhang","doi":"10.1016/j.eswa.2026.131168","DOIUrl":"10.1016/j.eswa.2026.131168","url":null,"abstract":"<div><div>Although existing studies have focused on efficiency promotion or threshold optimization in human-AI collaborative scoring for subjective assignments, there has been relatively little attention given to how changes in human capabilities and psychological states affect system performance. To fill this research gap, we present a novel simulation framework for human-AI collaborative scoring strategy from a view point of cognitive ergonomics. Firstly, we design a rater ability model with an associated score mapping function to simulate human performance under learning and fatigue effects. Secondly, we develop a three-objective optimization model to balance human costs, scoring fairness, and scoring consistency under the scenarios of learning and fatigue effects. Finally, considering decision-makers’ loss aversion psychology from behavioral economics, we utilize the behavioral technique for order of preference by similarity to ideal solution (TOPSIS) approach to rank human-AI collaboration schemes. In addition, performance comparisons and parametric analyses on eight prompts from the automated student assessment prize (ASAP) data set are conducted to demonstrate the feasibility and superiority of the proposed framework. More specifically, under the learning effect, an expert proportion of 6.39% leads to an average fairness gain of 62.93%, whereas under the fatigue effect, an expert proportion of 4.12% results in an average fairness gain of 61.07%. It is important to note that this work offers a conceptual simulation framework whose practical deployment necessitates empirical validation of its behavioral assumptions.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131168"},"PeriodicalIF":7.5,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980573","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

MSCF-net: Multi-scale frequency denoising and co-frequency enhancement network for multimodal recommendation MSCF-net：多模态推荐的多尺度频率去噪和共频增强网络

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-15 DOI: 10.1016/j.eswa.2026.131159

Chenghao Li , Wei Zhou , Yihao Zhang , Jiahao Hu , Huayi Shen , Junhao Wen

Multimodal Recommender Systems leverage rich side information to capture user preferences more accurately. However, existing methods often fuse different modalities via straightforward operators such as concatenation or attention, overlooking two critical challenges: i)The high-frequency components in features introduce modality-specific noise, which often carries low semantic relevance. ii)The sparse user-item interactions fail to capture vital co-frequency behavioral patterns. To address these issues, we propose MSCF-Net, a Multi-Scale Frequency Denoising and Co-Frequency Enhancement Network. Our key contribution lies in integrating frequency-domain insights throughout the recommendation process. We design an adaptive spectral domain filter with learnable weights to dynamically suppress high-frequency noise while preserving cross-modal semantics. Furthermore, we augment the bipartite interaction graph by constructing User-User and Item-Item Co-frequency Matrices derived from interaction histories, reinforcing collaborative signals through frequency-based affinity measurement. A novel Spectrum Harmonization Loss is also introduced to ensure both cross-modal alignment and modality-specificity preservation. These complementary mechanisms work together to better understand user preferences. On three real-world datasets, MSCF-Net shows encouraging performance compared to state-of-the-art baselines, validating the robustness and effectiveness of our proposed framework.

多模式推荐系统利用丰富的侧信息来更准确地捕获用户偏好。然而，现有的方法通常通过简单的操作符（如串联或注意）融合不同的模态，忽略了两个关键的挑战：i)特征中的高频成分引入了模态特定的噪声，这些噪声通常具有低语义相关性。ii)稀疏的用户-物品交互无法捕获重要的共频行为模式。为了解决这些问题，我们提出了MSCF-Net，一种多尺度频率去噪和共频增强网络。我们的主要贡献在于在整个推荐过程中集成了频域洞察力。设计了一种具有可学习权值的自适应谱域滤波器，在保持跨模态语义的同时动态抑制高频噪声。此外，我们通过构建基于交互历史的用户-用户和物品-物品共频矩阵来增强二部交互图，并通过基于频率的亲和度量来增强协作信号。还引入了一种新的频谱协调损失，以确保跨模态对齐和模态特异性保存。这些互补机制一起工作可以更好地理解用户偏好。在三个真实数据集上，与最先进的基线相比，MSCF-Net显示出令人鼓舞的性能，验证了我们提出的框架的鲁棒性和有效性。

{"title":"MSCF-net: Multi-scale frequency denoising and co-frequency enhancement network for multimodal recommendation","authors":"Chenghao Li , Wei Zhou , Yihao Zhang , Jiahao Hu , Huayi Shen , Junhao Wen","doi":"10.1016/j.eswa.2026.131159","DOIUrl":"10.1016/j.eswa.2026.131159","url":null,"abstract":"<div><div>Multimodal Recommender Systems leverage rich side information to capture user preferences more accurately. However, existing methods often fuse different modalities via straightforward operators such as concatenation or attention, overlooking two critical challenges: i)The high-frequency components in features introduce modality-specific noise, which often carries low semantic relevance. ii)The sparse user-item interactions fail to capture vital co-frequency behavioral patterns. To address these issues, we propose MSCF-Net, a <strong><u>M</u></strong>ulti-<strong><u>S</u></strong>cale Frequency Denoising and <strong><u>C</u></strong>o-<strong><u>F</u></strong>requency Enhancement Network. Our key contribution lies in integrating frequency-domain insights throughout the recommendation process. We design an adaptive spectral domain filter with learnable weights to dynamically suppress high-frequency noise while preserving cross-modal semantics. Furthermore, we augment the bipartite interaction graph by constructing User-User and Item-Item Co-frequency Matrices derived from interaction histories, reinforcing collaborative signals through frequency-based affinity measurement. A novel Spectrum Harmonization Loss is also introduced to ensure both cross-modal alignment and modality-specificity preservation. These complementary mechanisms work together to better understand user preferences. On three real-world datasets, MSCF-Net shows encouraging performance compared to state-of-the-art baselines, validating the robustness and effectiveness of our proposed framework.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"309 ","pages":"Article 131159"},"PeriodicalIF":7.5,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145981720","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

Fast heuristic search algorithms for submodular cost submodular cover under routing constraints 路由约束下子模代价子模覆盖的快速启发式搜索算法

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-14 DOI: 10.1016/j.eswa.2026.131134

Miao Yu , Xuefeng Chen , Shuang Liu , Liang Feng , Xin Cao , Zexuan Zhu

The problem of Submodular Cost Submodular Cover (SCSC) is to find a min-cost subset S from a set V such that the submodular benefit function exceeds a given threshold. The SCSC problem finds application in various real-world scenarios, and previous studies have primarily focused on easily solvable cost functions. However, many applications on the SCSC problem inherently involve complex cost functions, making exact cost computations difficult. This paper addresses this limitation by introducing the problem of SCSC under Routing Constraints (SCRC), which incorporates a typical complex cost function from route planning into the SCSC framework. The SCRC problem arises in diverse applications such as targeted mobile robotic sensing and targeted sales path planning. To tackle the formulated SCRC problem, we propose a cost-efficient greedy algorithm called CEGreedy and an acceleration technique to obtain solutions quickly. We also prove that CEGreedy has a better approximation guarantee compared to a commonly used greedy algorithm. Meanwhile, to escape from the local optima of CEGreedy, we devise an evolutionary algorithm named POSC, equipped with two schemes aimed at expediting the convergence process in practice. Finally, we validate the effectiveness and efficiency of the proposed algorithms and optimization approaches by conducting experiments on two applications and four real-world datasets.

子模成本-子模覆盖（Submodular Cost - Submodular Cover， SCSC）问题是从集合V中找到最小成本子集S，使得子模效益函数超过给定的阈值。SCSC问题在各种现实场景中都有应用，以前的研究主要集中在容易求解的成本函数上。然而，SCSC问题的许多应用固有地涉及复杂的成本函数，使得精确的成本计算变得困难。本文通过引入路由约束下的SCSC问题（SCRC）来解决这一限制，SCRC将路由规划中的典型复杂成本函数纳入SCSC框架。SCRC问题出现在各种应用中，如定向移动机器人传感和定向销售路径规划。为了解决公式化的SCRC问题，我们提出了一种成本高效的贪心算法CEGreedy和一种加速技术来快速获得解。我们还证明了CEGreedy算法比常用的贪心算法有更好的逼近保证。同时，为了摆脱CEGreedy的局部最优，我们设计了一种名为POSC的进化算法，该算法配备了两种方案，旨在加快实践中的收敛过程。最后，我们通过在两个应用程序和四个真实数据集上进行实验来验证所提出算法和优化方法的有效性和效率。

{"title":"Fast heuristic search algorithms for submodular cost submodular cover under routing constraints","authors":"Miao Yu , Xuefeng Chen , Shuang Liu , Liang Feng , Xin Cao , Zexuan Zhu","doi":"10.1016/j.eswa.2026.131134","DOIUrl":"10.1016/j.eswa.2026.131134","url":null,"abstract":"<div><div>The problem of Submodular Cost Submodular Cover (SCSC) is to find a min-cost subset <em>S</em> from a set <em>V</em> such that the submodular benefit function exceeds a given threshold. The SCSC problem finds application in various real-world scenarios, and previous studies have primarily focused on easily solvable cost functions. However, many applications on the SCSC problem inherently involve complex cost functions, making exact cost computations difficult. This paper addresses this limitation by introducing the problem of SCSC under Routing Constraints (SCRC), which incorporates a typical complex cost function from route planning into the SCSC framework. The SCRC problem arises in diverse applications such as targeted mobile robotic sensing and targeted sales path planning. To tackle the formulated SCRC problem, we propose a cost-efficient greedy algorithm called CEGreedy and an acceleration technique to obtain solutions quickly. We also prove that CEGreedy has a better approximation guarantee compared to a commonly used greedy algorithm. Meanwhile, to escape from the local optima of CEGreedy, we devise an evolutionary algorithm named POSC, equipped with two schemes aimed at expediting the convergence process in practice. Finally, we validate the effectiveness and efficiency of the proposed algorithms and optimization approaches by conducting experiments on two applications and four real-world datasets.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131134"},"PeriodicalIF":7.5,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980529","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 generalizable rapid multi-target capture framework based on a single conventional pan-tilt-zoom camera 一种基于单台传统泛倾斜变焦相机的通用快速多目标捕获框架

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-14 DOI: 10.1016/j.eswa.2026.131189

Fangxu Jiao , Minghao Zhang , Yi Tian , Yunlong Zhang , Deyun Ren , Yang Yang , Shan Zhao , Zidu Yin

In large-scale multi-target visual surveillance scenarios, multi-target object detection and fine-grained recognition are necessarily decoupled due to the limited clarity of wide-angle imaging. Pan-Tilt-Zoom (PTZ) cameras bridge coarse detection and fine-grained recognition, however, efficiently achieving high-quality transitions across multiple targets remains challenging. To address this, we propose a novel, generalizable rapid multi-target capture framework based on a single conventional PTZ camera (GRMCF), requiring no additional hardware. Central to our approach is formulating the rapid multi-target scheduling task as a Cost-driven 3D Traveling Salesman Problem with Overlapping Neighborhoods (C-3D-TSPON), incorporating non-convex viewing constraints and realistic PTZ motion time cost. Faced with an NP-hard path optimization problem in 3D space that jointly involves combinatorial sequencing and PTZ viewpoint selection, we further design a fast heuristic algorithm, OGSOA-GRO (Overlap-Aware Growing Self-Organizing Array enhanced with a Gradient-Based Refinement Operator), as an improved version of the Growing Self-Organizing Array (GSOA), which jointly optimizes the global capture sequence and per-target viewpoint configurations. It assigns a single high-quality viewpoint to multiple spatially adjacent targets when feasible, while maintaining sub-second solution time. Extensive synthetic and real-world experiments indicate that our method outperforms conventional strategies in terms of PTZ motion time cost, while also reducing the number of mechanical actuations, which is expected to mitigate long-term hardware stress, thereby achieving improved efficiency under strict imaging quality constraints. The proposed framework offers a scalable, hardware-compatible solution for intelligent multi-target visual acquisition in complex surveillance environments.

在大规模多目标视觉监控场景中，由于广角成像的清晰度有限，多目标检测和细粒度识别必须解耦。泛倾斜变焦（PTZ）相机桥粗检测和细粒度识别，然而，有效地实现高质量的过渡在多个目标仍然具有挑战性。为了解决这个问题，我们提出了一种基于单个传统PTZ相机（GRMCF）的新型，通用的快速多目标捕获框架，不需要额外的硬件。该方法的核心是将快速多目标调度任务制定为成本驱动的具有重叠邻域的3D旅行推销员问题（C-3D-TSPON），结合非凸观看约束和现实的PTZ运动时间成本。针对组合测序和PTZ视点选择共同涉及的三维空间NP-hard路径优化问题，我们进一步设计了一种快速启发式算法OGSOA-GRO（重叠感知生长自组织阵列增强与基于梯度的细化算子），作为生长自组织阵列（GSOA）的改进版本，共同优化全局捕获序列和每个目标的视点配置。它在可行的情况下为多个空间相邻目标分配一个高质量的视点，同时保持亚秒级的求解时间。大量的合成和实际实验表明，我们的方法在PTZ运动时间成本方面优于传统策略，同时还减少了机械驱动的数量，这有望减轻长期的硬件压力，从而在严格的成像质量限制下实现更高的效率。该框架为复杂监控环境下的智能多目标视觉采集提供了可扩展、硬件兼容的解决方案。

{"title":"A generalizable rapid multi-target capture framework based on a single conventional pan-tilt-zoom camera","authors":"Fangxu Jiao , Minghao Zhang , Yi Tian , Yunlong Zhang , Deyun Ren , Yang Yang , Shan Zhao , Zidu Yin","doi":"10.1016/j.eswa.2026.131189","DOIUrl":"10.1016/j.eswa.2026.131189","url":null,"abstract":"<div><div>In large-scale multi-target visual surveillance scenarios, multi-target object detection and fine-grained recognition are necessarily decoupled due to the limited clarity of wide-angle imaging. Pan-Tilt-Zoom (PTZ) cameras bridge coarse detection and fine-grained recognition, however, efficiently achieving high-quality transitions across multiple targets remains challenging. To address this, we propose a novel, generalizable rapid multi-target capture framework based on a single conventional PTZ camera (GRMCF), requiring no additional hardware. Central to our approach is formulating the rapid multi-target scheduling task as a Cost-driven 3D Traveling Salesman Problem with Overlapping Neighborhoods (C-3D-TSPON), incorporating non-convex viewing constraints and realistic PTZ motion time cost. Faced with an NP-hard path optimization problem in 3D space that jointly involves combinatorial sequencing and PTZ viewpoint selection, we further design a fast heuristic algorithm, OGSOA-GRO (Overlap-Aware Growing Self-Organizing Array enhanced with a Gradient-Based Refinement Operator), as an improved version of the Growing Self-Organizing Array (GSOA), which jointly optimizes the global capture sequence and per-target viewpoint configurations. It assigns a single high-quality viewpoint to multiple spatially adjacent targets when feasible, while maintaining sub-second solution time. Extensive synthetic and real-world experiments indicate that our method outperforms conventional strategies in terms of PTZ motion time cost, while also reducing the number of mechanical actuations, which is expected to mitigate long-term hardware stress, thereby achieving improved efficiency under strict imaging quality constraints. The proposed framework offers a scalable, hardware-compatible solution for intelligent multi-target visual acquisition in complex surveillance environments.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"309 ","pages":"Article 131189"},"PeriodicalIF":7.5,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145981721","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

Bi-objective home health care routing and scheduling problem based on caregiver-patient matching and satisfaction 基于照顾者与病患匹配与满意度的双目标家庭医疗照护路线与调度问题

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-14 DOI: 10.1016/j.eswa.2026.131138

Xianlong Ge, Hejing Ye, Qiushuang Yin, Mengdan Li

By the end of 2024, the proportion of the aging population over 65 years old has reached 10 % globally and is accelerating, which makes the problems of resource constraints and low service quality of the traditional healthcare model more prominent. Community hospitals and other healthcare organizations have adopted the home clinic service model, which provides convenient access, reduces costs, and enhances efficiency, leading to a greater recognition of its importance. Current research focuses on the impact of factors such as caregiver-patient matchmaking and patient satisfaction on the home healthcare service industry. Nonetheless, the study still neglects the mechanism of accurate dynamic matching optimization concerning the mixing time window of caregiver-patient and the satisfaction of caregivers under the high intensity of work. In this context, a two-stage mixed integer linear programming model is constructed based on dynamic matching of caregiver-patient preferences and patient priority, with satisfaction quantified using prospect theory. A hybrid heuristic algorithm has been proposed that combines the population coordination mechanism of the Ivy Algorithm (IVYA) with an improved third-generation non-dominated sorting genetic algorithm (NSGA-III). A large number of numerical experiments show that the algorithm significantly outperforms similar traditional algorithms in terms of Pareto solution set convergence and diversity on the Solomon benchmark dataset. The significant effects of the dynamic compensation system and the balanced healthcare workload on the path optimization scheme are demonstrated in an empirical study using Chongqing City as a case study. Therefore, our study provides valuable decision support for home healthcare service caregivers to make informed decisions and develop robust schedules for caregivers.

到2024年底，全球65岁以上老年人口比例已达10% %，并呈加速增长趋势，这使得传统医疗模式的资源约束和服务质量低下问题更加突出。社区医院和其他医疗机构已经采用了家庭诊所服务模式，这种模式提供了方便的访问，降低了成本，提高了效率，使人们更加认识到它的重要性。目前的研究主要集中在照顾者-病人配对、病人满意度等因素对家庭医疗保健服务行业的影响。然而，本研究仍然忽略了高工作强度下照顾者-患者混合时间窗与照顾者满意度的精确动态匹配优化机制。在此背景下，基于护理者偏好和患者优先级的动态匹配，构建了两阶段混合整数线性规划模型，并使用前景理论对满意度进行量化。将常青藤算法（IVYA）的种群协调机制与改进的第三代非支配排序遗传算法（NSGA-III）相结合，提出了一种混合启发式算法。大量的数值实验表明，在Solomon基准数据集上，该算法在Pareto解集收敛性和多样性方面明显优于类似的传统算法。以重庆市为例，实证研究了动态补偿制度和均衡医疗工作量对路径优化方案的显著影响。因此，我们的研究提供了有价值的决策支持，为家庭保健服务的护理人员做出明智的决策和制定健全的时间表。

{"title":"Bi-objective home health care routing and scheduling problem based on caregiver-patient matching and satisfaction","authors":"Xianlong Ge, Hejing Ye, Qiushuang Yin, Mengdan Li","doi":"10.1016/j.eswa.2026.131138","DOIUrl":"10.1016/j.eswa.2026.131138","url":null,"abstract":"<div><div>By the end of 2024, the proportion of the aging population over 65 years old has reached 10 % globally and is accelerating, which makes the problems of resource constraints and low service quality of the traditional healthcare model more prominent. Community hospitals and other healthcare organizations have adopted the home clinic service model, which provides convenient access, reduces costs, and enhances efficiency, leading to a greater recognition of its importance. Current research focuses on the impact of factors such as caregiver-patient matchmaking and patient satisfaction on the home healthcare service industry. Nonetheless, the study still neglects the mechanism of accurate dynamic matching optimization concerning the mixing time window of caregiver-patient and the satisfaction of caregivers under the high intensity of work. In this context, a two-stage mixed integer linear programming model is constructed based on dynamic matching of caregiver-patient preferences and patient priority, with satisfaction quantified using prospect theory. A hybrid heuristic algorithm has been proposed that combines the population coordination mechanism of the Ivy Algorithm (IVYA) with an improved third-generation non-dominated sorting genetic algorithm (NSGA-III). A large number of numerical experiments show that the algorithm significantly outperforms similar traditional algorithms in terms of Pareto solution set convergence and diversity on the Solomon benchmark dataset. The significant effects of the dynamic compensation system and the balanced healthcare workload on the path optimization scheme are demonstrated in an empirical study using Chongqing City as a case study. Therefore, our study provides valuable decision support for home healthcare service caregivers to make informed decisions and develop robust schedules for caregivers.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131138"},"PeriodicalIF":7.5,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980531","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

An augmented Lagrangian relaxation algorithm for allocation optimization of shared private parking spaces considering EV charging 考虑电动汽车充电的共享私人车位分配优化增广拉格朗日松弛算法

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-13 DOI: 10.1016/j.eswa.2026.131133

Jingyang Hou, Zhongkai Guo, Tong Li

As the ownership of fuel-powered and electric vehicles continues to grow, parking and charging difficulties have become particularly pronounced in densely populated areas. Shared private parking spaces, facilitated by shared parking platforms that activate idle private parking resources, offer an effective solution to alleviate these challenges. This paper investigates real-time allocation optimization for both conventional private parking resources and charging-capable private parking spaces within shared parking platforms, with the dual objectives of improving bilateral user satisfaction and enhancing parking space utilization efficiency, thereby mitigating the challenges of “parking difficulty” and “charging difficulty.” To this end, a shared parking allocation model based on a rolling horizon framework is developed, with the objective of maximizing bilateral satisfaction and parking space utilization. In addition, the platform incorporates two matching mechanisms to respond to different operational conditions. Given the NP-hard nature of the shared parking allocation problem, a heuristic algorithm based on Lagrangian relaxation and augmented Lagrangian relaxation is proposed to efficiently obtain near-optimal solutions. Numerical simulation results demonstrate that: (1) the designed matching rules effectively increase the number of successfully matched vehicles under various conditions; (2) compared with the commercial solver CPLEX, the proposed algorithm reduces computation time by 80.31%, with an average optimality gap of no more than 1.72%. In conclusion, the model and algorithm presented in this study show significant potential for optimizing resource allocation in shared parking platforms and enhancing social welfare.

随着燃油和电动汽车的拥有量不断增长，在人口密集地区，停车和充电的困难变得尤为明显。共享私人停车位，通过共享停车平台激活闲置的私人停车资源，为缓解这些挑战提供了有效的解决方案。本文以提高双方用户满意度和提高车位利用效率为双重目标，研究共享停车平台内传统私人停车资源和可收费私人停车位的实时分配优化，从而缓解“停车难”和“收费难”的挑战。为此，以双方满意度和车位利用率最大化为目标，建立了基于滚动地平线框架的共享车位分配模型。此外，该平台还采用了两种匹配机制来响应不同的操作条件。针对共享车位分配问题的NP-hard性质，提出了一种基于拉格朗日松弛和增广拉格朗日松弛的启发式算法，以有效地获得近最优解。数值仿真结果表明：(1)所设计的匹配规则有效地增加了各种条件下匹配成功的车辆数量；(2)与商用求解器CPLEX相比，本文算法的计算时间缩短了80.31%，平均最优性差距不超过1.72%。综上所述，本文提出的模型和算法在优化共享停车平台资源配置和提高社会福利方面具有重要的潜力。

{"title":"An augmented Lagrangian relaxation algorithm for allocation optimization of shared private parking spaces considering EV charging","authors":"Jingyang Hou, Zhongkai Guo, Tong Li","doi":"10.1016/j.eswa.2026.131133","DOIUrl":"10.1016/j.eswa.2026.131133","url":null,"abstract":"<div><div>As the ownership of fuel-powered and electric vehicles continues to grow, parking and charging difficulties have become particularly pronounced in densely populated areas. Shared private parking spaces, facilitated by shared parking platforms that activate idle private parking resources, offer an effective solution to alleviate these challenges. This paper investigates real-time allocation optimization for both conventional private parking resources and charging-capable private parking spaces within shared parking platforms, with the dual objectives of improving bilateral user satisfaction and enhancing parking space utilization efficiency, thereby mitigating the challenges of “parking difficulty” and “charging difficulty.” To this end, a shared parking allocation model based on a rolling horizon framework is developed, with the objective of maximizing bilateral satisfaction and parking space utilization. In addition, the platform incorporates two matching mechanisms to respond to different operational conditions. Given the NP-hard nature of the shared parking allocation problem, a heuristic algorithm based on Lagrangian relaxation and augmented Lagrangian relaxation is proposed to efficiently obtain near-optimal solutions. Numerical simulation results demonstrate that: (1) the designed matching rules effectively increase the number of successfully matched vehicles under various conditions; (2) compared with the commercial solver CPLEX, the proposed algorithm reduces computation time by 80.31%, with an average optimality gap of no more than 1.72%. In conclusion, the model and algorithm presented in this study show significant potential for optimizing resource allocation in shared parking platforms and enhancing social welfare.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131133"},"PeriodicalIF":7.5,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980721","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

An evolutionary multitasking with elbow principal component analysis and negative transfer optimization for high-dimensional feature selection 基于肘主成分分析和负传递优化的高维特征选择进化多任务

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-13 DOI: 10.1016/j.eswa.2026.131139

Jiayi Wang , Yujia Wang , Xiaoyu Su , Jingzhi Zhang , Cancan Liu

In recent years, evolutionary multitasking (EMT) methods have demonstrated significant advantages in addressing feature selection (FS) problems. However, the feature redundancy and negative transfer happened when auxiliary task construct and knowledge transfer in most of EMT-based FS methods. To reduce feature redundancy and negative transfer, this paper proposes an evolutionary multitasking with elbow principal component analysis and negative transfer optimization for high-dimensional feature selection (PN-EMTFS). PN-EMTFS integrates an elbow principal component analysis (PCA) and a negative transfer optimization strategy, thereby enhancing the performance of high-dimensional FS algorithms. The proposed algorithms are evaluated on 26 high-dimensional datasets ranging from 2308 to 22,883 dimensions. Notably, it excelled comparative algorithms in 20 datasets concerning classification accuracy, 8 datasets concerning the number of selected features, and 14 datasets concerning CPU runtime out of the 26 datasets considered. The results showed that PN-EMTFS performed at performance levels exceeding 90% for F1-score,sensitivity and sensitivity metrics, respectively.

近年来，进化多任务（EMT）方法在解决特征选择（FS）问题方面显示出显著的优势。然而，大多数基于emt的FS方法在辅助任务构建和知识转移过程中存在特征冗余和负迁移问题。为了减少特征冗余和负迁移，提出了一种基于肘主成分分析和负迁移优化的进化多任务高维特征选择方法（PN-EMTFS）。PN-EMTFS集成了肘主成分分析（PCA）和负传递优化策略，从而提高了高维FS算法的性能。在2308 ~ 22,883维的26个高维数据集上对所提算法进行了评估。值得注意的是，在考虑的26个数据集中，它在20个分类精度数据集、8个选择特征数量数据集和14个CPU运行时间数据集上优于比较算法。结果表明，PN-EMTFS在f1评分、灵敏度和灵敏度指标上的性能水平分别超过90%。

{"title":"An evolutionary multitasking with elbow principal component analysis and negative transfer optimization for high-dimensional feature selection","authors":"Jiayi Wang , Yujia Wang , Xiaoyu Su , Jingzhi Zhang , Cancan Liu","doi":"10.1016/j.eswa.2026.131139","DOIUrl":"10.1016/j.eswa.2026.131139","url":null,"abstract":"<div><div>In recent years, evolutionary multitasking (EMT) methods have demonstrated significant advantages in addressing feature selection (FS) problems. However, the feature redundancy and negative transfer happened when auxiliary task construct and knowledge transfer in most of EMT-based FS methods. To reduce feature redundancy and negative transfer, this paper proposes an evolutionary multitasking with elbow principal component analysis and negative transfer optimization for high-dimensional feature selection (PN-EMTFS). PN-EMTFS integrates an elbow principal component analysis (PCA) and a negative transfer optimization strategy, thereby enhancing the performance of high-dimensional FS algorithms. The proposed algorithms are evaluated on 26 high-dimensional datasets ranging from 2308 to 22,883 dimensions. Notably, it excelled comparative algorithms in 20 datasets concerning classification accuracy, 8 datasets concerning the number of selected features, and 14 datasets concerning CPU runtime out of the 26 datasets considered. The results showed that PN-EMTFS performed at performance levels exceeding 90% for F1-score,sensitivity and sensitivity metrics, respectively.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131139"},"PeriodicalIF":7.5,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980526","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

Information entropy based evolutionary multitasking optimization 基于信息熵的进化多任务优化

IF 7.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Expert Systems with Applications

Pub Date : 2026-01-13 DOI: 10.1016/j.eswa.2026.131137

Shuijia Li , Rui Wang , Wenyin Gong , Yanchi Li , Delong Chen , Zuowen Liao

Evolutionary multitasking optimization (EMTO) has emerged as a promising paradigm that simultaneously solves multiple optimization tasks by leveraging potential inter-task similarities. However, existing approaches often suffer from negative knowledge transfer due to overdependence on elite solutions. To alleviate this issue, in this paper, an information entropy based evolutionary multitasking differential evolution algorithm referred as IEDE, was proposed. The novelty of IEDE lies primarily in the proposed information entropy guidance mechanism, which is specifically reflected in the following two points: 1) an adaptive mutation strategy selection mechanism based on information entropy, which effectively balances exploration and exploitation by dynamically switching between DE/rand/1 and DE/best/1 operators according to the real-time population diversity, and 2) a knowledge transfer strategy based on information entropy, which selects whether to use a source individual’s knowledge for transfer not only based on their fitness value but also according to the amount of information entropy they carry for the target individual. To validate the effectiveness of the proposed IEDE, three multi-task optimization benchmark problems, CEC17, WCCI20, and CMT, were selected for the test suite. The comprehensive experimental results show that the proposed method achieves better or competitive performance compared to several well-established EMTO algorithms. Furthermore, IEDE has also demonstrated its considerable potential in two real-world application problems.

进化多任务优化（EMTO）已经成为一种很有前途的范例，它通过利用潜在的任务间相似性同时解决多个优化任务。然而，由于过度依赖精英解决方案，现有方法往往存在负知识转移的问题。为了解决这一问题，本文提出了一种基于信息熵的多任务进化差分进化算法。IEDE的新颖性主要在于提出的信息熵引导机制，具体体现在以下两点：1)基于信息熵的自适应突变策略选择机制，根据实时种群多样性动态切换DE/rand/1和DE/best/1算子，有效平衡探索和开发；2)基于信息熵的知识转移策略；它不仅根据源个体的适应度值来选择是否使用源个体的知识进行迁移，而且还根据源个体对目标个体所携带的信息熵的大小来选择是否使用源个体的知识进行迁移。为了验证所提出的IEDE的有效性，我们选择了三个多任务优化基准问题CEC17、WCCI20和CMT作为测试套件。综合实验结果表明，与几种已有的EMTO算法相比，该方法具有更好的性能或具有竞争力。此外，IEDE还在两个实际应用问题中展示了其相当大的潜力。

{"title":"Information entropy based evolutionary multitasking optimization","authors":"Shuijia Li , Rui Wang , Wenyin Gong , Yanchi Li , Delong Chen , Zuowen Liao","doi":"10.1016/j.eswa.2026.131137","DOIUrl":"10.1016/j.eswa.2026.131137","url":null,"abstract":"<div><div>Evolutionary multitasking optimization (EMTO) has emerged as a promising paradigm that simultaneously solves multiple optimization tasks by leveraging potential inter-task similarities. However, existing approaches often suffer from negative knowledge transfer due to overdependence on elite solutions. To alleviate this issue, in this paper, an information entropy based evolutionary multitasking differential evolution algorithm referred as IEDE, was proposed. The novelty of IEDE lies primarily in the proposed information entropy guidance mechanism, which is specifically reflected in the following two points: 1) an adaptive mutation strategy selection mechanism based on information entropy, which effectively balances exploration and exploitation by dynamically switching between DE/rand/1 and DE/best/1 operators according to the real-time population diversity, and 2) a knowledge transfer strategy based on information entropy, which selects whether to use a source individual’s knowledge for transfer not only based on their fitness value but also according to the amount of information entropy they carry for the target individual. To validate the effectiveness of the proposed IEDE, three multi-task optimization benchmark problems, CEC17, WCCI20, and CMT, were selected for the test suite. The comprehensive experimental results show that the proposed method achieves better or competitive performance compared to several well-established EMTO algorithms. Furthermore, IEDE has also demonstrated its considerable potential in two real-world application problems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"308 ","pages":"Article 131137"},"PeriodicalIF":7.5,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980399","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