Expert Systems with Applications最新文献

{"title":"EMOCSO: an efficient multi-objective competitive swarm optimizer for large-scale optimization","authors":"Wuxin Li ,&nbsp;Jie Yang ,&nbsp;Huiduo Wang ,&nbsp;Yanhong Wang","doi":"10.1016/j.eswa.2025.130060","DOIUrl":"10.1016/j.eswa.2025.130060","url":null,"abstract":"<div><div>Large-scale multi-objective optimization problems (LSMOPs) are crucial in real-world applications where balancing conflicting objectives is essential for decision-making. To overcome this limitation, we propose an efficient Multi-objective Competitive Swarm Optimizer (EMOCSO). The algorithm introduces three key innovations: (1) an Archive-driven Winner Learning Strategy that uses elite solutions to guide the search, (2) a Dual-layer Differential Neutral Update Mechanism to enhance diversity by adaptively updating “neutral” individuals (solutions with similar fitness), and (3) a Selective Spiral Archive Update Strategy to refine solutions through spiral-based local search. Comprehensive experiments on the LSMOP benchmark show that EMOCSO outperforms five state-of-the-art algorithms, demonstrating superior convergence and diversity in high-dimensional optimization scenarios. Moreover, in the application of active power allocation for regional photovoltaic clusters driven by meteorological data, EMOCSO effectively balances multiple objectives such as following dispatch instructions, stabilizing power output, and controlling uncertainty, providing operable solutions for actual power grid dispatching and verifying its engineering practical value.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130060"},"PeriodicalIF":7.5,"publicationDate":"2025-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364849","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}

{"title":"Dynamic link prediction in construction innovation networks: An integrated framework of topological and content attributes","authors":"Yajiao Chen ,&nbsp;Qinghua He ,&nbsp;Xiaoyan Chen ,&nbsp;Likai Zheng","doi":"10.1016/j.eswa.2025.130061","DOIUrl":"10.1016/j.eswa.2025.130061","url":null,"abstract":"<div><div>Accurately predicting the evolution of collaborative relationships in construction innovation organizations is crucial for optimizing subsequent innovation decisions and developing inter-organizational collaboration strategies. Regrettably, prior research has accorded limited attention to link prediction within construction innovation collaborative networks. Consequently, this study introduces a novel link prediction approach that forecasts inter-node connectivity relationships by integrating topological structure characteristics and node content attributes of these networks. The proposed metrics leverage primary and secondary measures based on temporal events to assess the influence of nodes and their neighbors on the prediction outcomes. Through a comprehensive set of experiments, the study systematically assessed the performance of the proposed metrics across diverse link prediction scenarios. The results indicate that the proposed metric consistently outperforms several well-established baseline methods, yielding highly encouraging outcomes. This research not only enriches the theoretical underpinnings of network link prediction and construction innovations but also provides valuable insights into the evolution of collaborative relationships and the identification of potential partners within innovation organizations.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130061"},"PeriodicalIF":7.5,"publicationDate":"2025-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364834","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}

{"title":"Negative samples filter of contrastive learning for time series classification","authors":"Yinlong Li,&nbsp;Licheng Pan,&nbsp;Hu Xu,&nbsp;Xinggao Liu","doi":"10.1016/j.eswa.2025.130052","DOIUrl":"10.1016/j.eswa.2025.130052","url":null,"abstract":"<div><div>As an unsupervised learning method, contrastive learning has achieved remarkable success in the field of computer vision. However, issues such as false negative samples and hard negative samples can significantly impair its effectiveness. Addressing these issues is therefore crucial for improving contrastive learning. While current research on handling these challenges mainly focuses on image data, there is limited exploration of contrastive learning for time series data. In this paper, we propose a negative samples filter in the embedding space to investigate the impact of hard negative samples on time series contrastive learning. We conducted extensive experiments on six different time series datasets to examine the effect of the negative samples filter on classification performance, both in unsupervised and supervised settings. The results demonstrate that in the unsupervised case, some of the most difficult samples can degrade classification performance, while in the supervised case, more difficult samples are beneficial for classification. Furthermore, we applied our filter function to other contrastive learning baselines for time series, achieving superior results compared to previous baselines, and outperforming other baselines that address false negative and hard negative samples.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130052"},"PeriodicalIF":7.5,"publicationDate":"2025-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364833","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}

{"title":"Achieving dynamic controllability for simple temporal networks with uncertainty and sensing timepoints","authors":"Xianzhang Cheng ,&nbsp;Chao Qi ,&nbsp;Hongwei Wang ,&nbsp;Yuhui Gao","doi":"10.1016/j.eswa.2025.130009","DOIUrl":"10.1016/j.eswa.2025.130009","url":null,"abstract":"<div><div>With the widespread deployment of advanced sensors, executors in real-world planning domains can acquire information about temporal uncertainty through sensing activities, enabling the resolution of previously intractable planning problems. This development underscores the necessity of finding more dynamically controllable plans by leveraging sensing activities that reduce temporal uncertainty, thereby enhancing the solvability of planning problems under uncertainty. This paper addresses the problem of transforming weakly controllable temporal plans into dynamically controllable ones by inserting a minimal set of sensing activities. We propose Simple Temporal Network with Uncertainty and Sensing Timepoints (STNUST), an extension of the traditional Simple Temporal Network with Uncertainty (STNU) model that explicitly incorporates sensing activities. To support this model, we develop ST-BOSA, a novel algorithm composed of four interdependent modules for constraint propagation, redundancy elimination, sensing timepoint selection, and insertion. Extensive experiments on Mars rover-inspired scenarios and randomly generated networks demonstrate that the proposed approach effectively achieves dynamic controllability for networks while minimizing the number of inserted sensing timepoints. This framework shows promise for integration into planning systems in sensor-rich domains such as space exploration. Future work includes improving scalability and extending support to multi-agent settings.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130009"},"PeriodicalIF":7.5,"publicationDate":"2025-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364855","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}

{"title":"Dynamic feature fusion guiding and multimodal large language model refining for medical image report generation","authors":"Pu Han ,&nbsp;Xiong Li ,&nbsp;Shenqi Jing ,&nbsp;Jianxiang Wei","doi":"10.1016/j.eswa.2025.130082","DOIUrl":"10.1016/j.eswa.2025.130082","url":null,"abstract":"<div><div>Medical image report generation refers to the automatic generation of text descriptions that correspond to specific medical images. In recent years, the increasing demand for medical imaging from both patients and healthcare institutions has significantly increased radiologists’ workloads. Concurrently, shortages in medical resources and diagnostic capabilities have raised the risks of diagnostic delays and misinterpretations in medical imaging. To alleviate the burden on medical professionals and ensure accurate diagnoses, the task of automated medical report generation has attracted a growing number of researchers. In this context, systems based on deep learning methods combined with general Large Language Models (LLMs) have been developed. However, existing methods face limitations in effectively integrating visual and textual data and they ignore the fact that the contributions of different modalities to diagnostic results vary across cases. Additionally, these approaches fail to address the lack of specialized medical knowledge when applying general LLMs. This paper introduces the Dynamic Feature Fusion Guiding and Multimodal Large Language Model Refining (DFFG-MLLMR) framework, which addresses these limitations through two key components:(1) The DFFG module dynamically adjusts the contributions of visual and textual features based on their diagnostic relevance, ensuring optimal feature utilization for report generation; (2) The MLLMR module integrates visual retrieval methods with fine-tuned LLMs to generate comprehensive and accurate medical reports. Our method achieves quantitatively superior results to other baseline methods on both benchmark datasets. On the IU-Xray dataset, DFFG-MLLMR achieves BLEU-4 of 0.191 and CIDEr of 0.574, exceeding the best conventional approach Token-Mixer. On the MIMIC-CXR dataset, our method achieves BLEU-4 of 0.132 and CIDEr of 0.289, improving upon Token-Mixer by 0.008 and 0.126. Experiments on public datasets demonstrate the superiority of DFFG-MLLMR, showing significant improvements in cross-modal feature fusion performance and enhanced diagnostic quality in automated reports. Furthermore, ablation studies confirm that the DFFG and MLLMR modules contribute complementary improvements, collectively enhancing the accuracy and clinical reliability of reports. The code can be obtained at <span><span>https://github.com/BearLiX/DFFG-MLLMR</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130082"},"PeriodicalIF":7.5,"publicationDate":"2025-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364831","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}

{"title":"FreLinear: spectral-aware design and acceleration for efficient graph neural networks","authors":"Xuefeng Li ,&nbsp;Zhengyuan Wang ,&nbsp;Chensu Zhao ,&nbsp;Xiaqiong Fan ,&nbsp;Xinxin Zhang ,&nbsp;Honglin Xie","doi":"10.1016/j.eswa.2025.130066","DOIUrl":"10.1016/j.eswa.2025.130066","url":null,"abstract":"<div><div>Graph Neural Networks (GNNs) excel in modeling graph-structured data but often face significant computational costs and fail to capture high-frequency components critical for fine-grained local variations. We propose FreLinear, a novel framework that integrates spectral-domain analysis with an efficient linear-attention mechanism. By avoiding the quadratic complexity inherent in traditional Transformer architectures, FreLinear leverages Fourier-based spectral features to enhance sensitivity to local structures while achieving near-linear computational complexity. Extensive experiments across diverse benchmark datasets demonstrate that FreLinear consistently surpasses state-of-the-art GNNs, delivering superior accuracy with significantly reduced computational overhead. On eight public datasets such as arxiv and Citeseer, the running time was shortened by 1 to 3 times with an increase in the number of parameters. At the same time, on the node classification task, the performance was improved by an average of 1.4 percentage points compared to the previous best work in these eight datasets. The code for the method proposed in our paper is publicly available on <span><span>https://github.com/SWLee777/Frelinear</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130066"},"PeriodicalIF":7.5,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364838","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}

{"title":"Cascade-TCN-BiLSTM: accurate prediction of long-term transmission error curves in multi-stage transmission system","authors":"Xiao Wang ,&nbsp;Hao Gong ,&nbsp;Jianhua Liu ,&nbsp;Ruixiang Wang ,&nbsp;Zhongtian Lu","doi":"10.1016/j.eswa.2025.130023","DOIUrl":"10.1016/j.eswa.2025.130023","url":null,"abstract":"<div><div>Accurately forecasting long-term transmission error trends in multi-stage transmission systems is essential for ensuring high motion accuracy in mechanical systems. Effectively modeling the nonlinear propagation and inter-stage coupling of errors to enhance predictive capabilities remains a significant challenge. This research introduces a cascaded deep learning framework, termed Cascade-Temporal Convolutional Network-Bidirectional Long Short-Term Memory, designed to estimate long-term transmission error curves across planetary and harmonic stages. By building a three-stage cascade aligned with intrinsic errors of the planetary reducer, inter-stage assembly errors at the planetary–harmonic interface, and operational errors of the harmonic reducer, we establish a one-to-one mapping between network modules and the corresponding error sources, thereby ensuring physical interpretability. The model incorporates both static assembly features and short-term dynamic input signals. A stage-specific cascaded configuration is embedded into a comprehensive sequence-to-sequence structure, consisting of an encoder-decoder network. Each encoder and decoder component consists of stacked temporal convolutional networks and bidirectional long short-term memory layers, followed by a multi-head attention module designed. Experimental results indicate that the proposed model consistently achieves low mean squared error and mean absolute error, typically below 0.22 and 0.33, respectively. The coefficient of determination exceeds 0.97 in most cases, demonstrating that the model significantly outperforms both traditional machine learning methods and baseline deep learning architectures. Ablation studies further confirm the critical contributions of the unified architecture, temporal modeling, and attention mechanism to the model’s performance. In addition to multi-stage transmissions, the method applies to series elastic actuators, surgical and industrial robot joints, and rotating machinery.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130023"},"PeriodicalIF":7.5,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364836","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}

{"title":"Bridging the safety-specific language model gap: Domain-adaptive pretraining of transformer-based models across several industrial sectors for occupational safety applications","authors":"Abid Ali Khan Danish,&nbsp;Snehamoy Chatterjee","doi":"10.1016/j.eswa.2025.130068","DOIUrl":"10.1016/j.eswa.2025.130068","url":null,"abstract":"<div><div>Occupational safety remains a persistent global challenge despite advancements in regulatory frameworks and safety technologies. Unstructured incident narratives, such as accident reports and safety logs, offer valuable context for understanding workplace hazards but are underutilized due to the gap in the safety-specific language models. This study addresses that gap by adapting pretrained transformer-based models (BERT and ALBERT) to the occupational safety domain through Domain-Adaptive Pretraining (DAPT). We construct a large-scale, multi-source corpus comprising over 2.4 million documents spanning several industrial sectors, including mining, construction, transportation, and chemical processing, augmented with safety-related academic abstracts to preserve general linguistic understanding and mitigate catastrophic forgetting. Using this corpus, we develop two domain-adapted models, safetyBERT and safetyALBERT, through continual pretraining on the masked language modeling objective. Intrinsic evaluation using pseudo-perplexity (PPPL) demonstrates substantial improvements, with safetyBERT and safetyALBERT achieving 76.9% and 90.3% reductions in PPPL, respectively, over their general-domain counterparts. Extrinsic evaluation on the Mine Safety and Health Administration (MSHA) injury dataset across three classification tasks (accident type, mining equipment, and degree of injury) demonstrated consistent performance improvements, with both models outperforming diverse baseline models including general-purpose models (BERT, ALBERT, DistilBERT, RoBERTa), domain-specific scientific model (SciBERT), and large language model (Llama 3.1-8B), with safetyALBERT achieving competitive results despite its parameter-efficient design.. To further assess generalization in low-resource settings, these models were evaluated on the small-scale Alaska insurance claim dataset from mining industry across two classification tasks − claim type and injured body part. Both safetyBERT and safetyALBERT maintained strong performance under this constraint, demonstrating the value of domain adaptation for data-constrained environments. Additionally, multi-task classification on the MSHA dataset using safety domain models showed improved generalization and more balanced performance across underrepresented classes. These findings confirm that DAPT effectively enhances language understanding in safety–critical domains while enabling scalable, resource-efficient deployment. This work lays the foundation for integrating domain-adapted natural language processing (NLP) systems into occupational health and safety management frameworks.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130068"},"PeriodicalIF":7.5,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364832","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}

{"title":"A Comprehensive Framework for Human - AI Collaborative Decision Making in Intelligent Retail Environments","authors":"Sunaina Sridhar ,&nbsp;Praveen Baskar ,&nbsp;Josh Grimes ,&nbsp;Ashwin Sampathkumar","doi":"10.1016/j.eswa.2025.130013","DOIUrl":"10.1016/j.eswa.2025.130013","url":null,"abstract":"<div><div>Artificial intelligence (AI) approaches have been more and more adopted in the retail industry in the past years, ranging from demand forecasting, dynamic pricing, inventory optimization to personalization of recommendations and promotions. However, conventional AI-centric decision platforms are often limited in interpretability, unable to manage data heterogeneity across channels, real-time adaptability and lack of domain knowledge from human expertise. Intelligent retailing is one application field that this paper would propose a human-AI cooperative decision-making system in order to combine the benefits of human expertise and machine learning. This system should be developed on: (i) modular architecture that includes a reinforcement learning (RL) core, fuzzy logic reasoning engine, human feedback interface, bias detection module; (ii) explainable AI (XAI) methods to output the rationale of the model, and also have human operators for (iii) human-in-the-loop correction and (iv) bias mitigation and fairness checks, and (v) a hybrid multi-store evaluation mechanism. Experiment: we compare our framework against baselines such as traditional rule-based systems, pure RL models and the more recent hybrid human-AI methods. Experiments are based on six months of transaction and inventory data from three separate mid-size retail stores (&gt; 500,000 transactions, ∼2,000 SKUs), with results showing an increase of 15 percent in revenue and 10–12 percent reduction in stock-outs, and an average increase of around 18 percent in staff satisfaction indices, and with decision latency below 200 ms. The advantage can be shown by paired t-tests (ANOVA, p = 0.05). Ablation experiments demonstrate the importance of each of the modules (e.g., XAI transparency, fuzzy logic smoothing, bias detector). The qualitative interview data with store managers on the explanations and override controls provide a basis for trust.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130013"},"PeriodicalIF":7.5,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364841","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}

{"title":"An efficient column generation approach for crew re-scheduling and recovery in urban rail transit systems under emergency conditions","authors":"Mengjiao Zhao ,&nbsp;Songpo Yang ,&nbsp;Xin Yang ,&nbsp;Jianjun Wu","doi":"10.1016/j.eswa.2025.129993","DOIUrl":"10.1016/j.eswa.2025.129993","url":null,"abstract":"<div><div>Crew Re-Scheduling Problem is a significant challenge in urban rail transit systems, particularly when addressing service disruptions and restoring operational order. When crew members unexpectedly sign off due to emergencies (e.g., illness), the train assigned to their operation task may be stranded in one running direction. This can subsequently cause obstructions for trains following in the same direction, thereby impacting normal operations. To address this issue, we first propose introducing a closed-loop scheduling mode, which involves rearranging the finite crew members across both running directions to sustain operations during emergency periods. Subsequently, a Crew Re-Scheduling and Recovery (CRSRP) model is developed to response the depart-time changes of trains. To solve the model, a generic framework of column generation (CG) embedded labeling algorithm is re-engineered to meet re-scheduling time requirements and permit changes in running directions at disrupted stations, which could be adopted in different emergency phases. It is important to note that after fireman crews are supplemented, all crew members resume normal operations, but emergency tasks must still be prioritized. A greedy algorithm is devised to manage assignments during the recovery phase. Finally, a real-life case study from Beijing is presented to assess the effectiveness of the proposed method. The model demonstrates the capability to respond swiftly within 30 min post-accident and control the generation time of individual tasks within 1 min. Additionally, the fluctuation range of crew members’ scheduling time has been reduced to [4, 21] minutes. This evidence underscores the model’s efficacy in restoring operational order under emergency conditions.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 129993"},"PeriodicalIF":7.5,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145334035","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}