Expert Systems with Applications最新文献

{"title":"Adaptive Multi-Criteria group Decision-Making for optimized crop selection using the p,q- quasirung orthopair fuzzy hybrid aggregation operator","authors":"Muhammad Rahim ,&nbsp;Kamal Shah ,&nbsp;Haifa Alqahtani ,&nbsp;Somayah Abdualziz Alhabeeb ,&nbsp;Hamiden Abd El-Wahed Khalifa","doi":"10.1016/j.eswa.2025.128126","DOIUrl":"10.1016/j.eswa.2025.128126","url":null,"abstract":"<div><div>This study introduces the <span><math><mrow><mi>p</mi><mo>,</mo><mi>q</mi><mo>-</mo></mrow></math></span> Quasirung Orthopair Fuzzy Hybrid Aggregation (<span><math><mrow><mi>p</mi><mo>,</mo><mi>q</mi><mo>-</mo></mrow></math></span> QOFHA) operator, designed to enhance multi-criteria group decision-making (MCGDM) under uncertainty and imprecise information. Unlike traditional aggregation operators, which lack flexibility in handling dynamic environmental factors, the proposed <span><math><mrow><mi>p</mi><mo>,</mo><mi>q</mi><mo>-</mo></mrow></math></span> QOFHA operator incorporates two adjustable parameters, <span><math><mi>p</mi></math></span> and <span><math><mi>q</mi></math></span>, enabling adaptive control over decision-making. These parameters allow decision-makers to account for environmental changes, weather conditions, and external factors affecting agricultural productivity, making the approach more robust and practical for real-world applications. The study applies the proposed method to an optimized crop selection problem, evaluating seven crop alternatives (wheat, rice, maize, sugarcane, soybean, barley, and cotton) based on five critical criteria (soil fertility, water availability, temperature, market demand, and sustainability). The entropy-based weighting approach is used to determine criteria importance. Unlike traditional fuzzy MCDM models, which treat uncertainty rigidly, the <span><math><mrow><mi>p</mi><mo>,</mo><mi>q</mi><mo>-</mo></mrow></math></span> QOFHA operator dynamically adjusts membership and non-membership values, ensuring a more realistic and flexible decision-making process. Sensitivity analysis further demonstrates that adjusting <span><math><mi>p</mi></math></span> and <span><math><mi>q</mi></math></span> enables for adaptive responses to changing agricultural conditions, outperforming conventional fuzzy aggregation approaches. The findings highlight the superiority of the proposed <span><math><mrow><mi>p</mi><mo>,</mo><mi>q</mi><mo>-</mo></mrow></math></span> QOFHA operator in handling uncertainty and adapting to external influences, making it a powerful tool for sustainable agricultural decision-making. This study provides a foundation for applying advanced fuzzy MCDM models in climate adaptation, resource management, and other complex decision-making scenarios where dynamic environmental factors play a crucial role.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128126"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144069140","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":"COO-DuDo: computation overhead optimization methods for dual-domain sparse-view CT reconstruction","authors":"Zihan Deng ,&nbsp;Zhisheng Wang ,&nbsp;Yuanlin Shan ,&nbsp;Guohang He ,&nbsp;Tiantian Du ,&nbsp;Shunli Wang","doi":"10.1016/j.eswa.2025.128109","DOIUrl":"10.1016/j.eswa.2025.128109","url":null,"abstract":"<div><div>Recently, deep learning methods have shown exciting effects in Sparse-view CT reconstruction. The Dual-Domain (DuDo) deep learning method is one of the representative methods, and it can process the information in both the sinogram and image domains. However, the existing DuDo methods do not pay enough attention to the allocation of training costs and strategies for the two domains, which will result in wasted computing overhead or insufficient training in one of the two domains. In this paper, we propose a Computation-Overhead Optimization (COO) DuDo training strategy for sparse-view CT reconstruction, i.e., COO-DuDo. The training ratio of different domains is controlled by calculating their computation overhead, loss, and gradient variation of the loss. To make our COO-DuDo strategy enable sparse-view CT reconstruction better, we adopt a DuDo-Network (COO-DDNet) structure based on two coding-decoding-type subnetworks. The evaluation, animal and clinical experiments have verified the effectiveness of our training strategies and methods. Our research provides a broader perspective for dual-domain image restoration tasks from the perspective of computational overhead.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128109"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143947817","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":"Towards efficient solutions for automatic recognition of complex power quality disturbances","authors":"Abderrezak Laouafi","doi":"10.1016/j.eswa.2025.128077","DOIUrl":"10.1016/j.eswa.2025.128077","url":null,"abstract":"<div><div>The integration of renewable energy sources and the emergence of many innovative technologies make abnormal deviations in voltage waveforms more complex and severe, as different combinations of power quality disturbances (PQDs) are likely to be produced simultaneously, significantly impacting the reliability, security, and stability of the grid. Unlike previous studies that considered only a small number of single and double PQDs, the present research addresses the challenge of identifying multiple PQDs with superimposition of up to 4 single disturbances on the same waveform. To this end, a new model is proposed in this paper, which combines the principles of wavelet denoising, hybrid signal processing, feature selection, and pattern classification with a bagged ensemble of decision trees. The main idea behind this integration is to enhance information diversity, track the amplitude variation of complex PQDs, and achieve better generalization capability while ensuring a trade-off between accuracy and computational efficiency. Due to the lack of reliable data on power quality studies, open-source software and a synthetic dataset containing 71 types of disturbances are also provided to support future work and serve as references for evaluating and comparing different methods. The results obtained by the study show: (1) an accuracy rate of 97.03 %, 96.82 %, 96.60 % and 95.16 % for noise-free, 50 dB, 40 dB and 30 dB SNR cases, respectively; (2) superior performance compared to 28 state-of-the-art algorithms; (3) average computation time of 0.5779 s; and (4) promising potential for recognizing PQDs with a large number of possible classes.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128077"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143948416","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":"Adaptive Kalman filter-based real-time distribution estimation of unmeasured system states for an advanced passive reactor core under optimized mechanical shim control strategy","authors":"Jiuwu Hui","doi":"10.1016/j.eswa.2025.128006","DOIUrl":"10.1016/j.eswa.2025.128006","url":null,"abstract":"<div><div>Reactor cores (RCs) serve as the heart of modern nuclear power plants (NPPs), necessicating efficient and reliabe monitoring during the load following operation. However, most of the critical system states within the RC, such as reactivity, fuel temperature, and xenon concentration, cannot be measured directly owing to safety reasons and technical restrictions; moreover, the RC exhibits extremely nonlinear and time-varying dynamics, compounded by the unknown process and measurement noises, posing significant challenges to advanced estimator design and implementation. Especially for the axial distribution estimation of these unmeasured system states, the relevant studies and reports are rare in the published literature. To this end, this paper is dedicated to achieving the real-time distribution estimation of unmeasured system states for an advanced passive reactor (AP1000)-type RC across its full operating range. A novel estimation algorithm integrating the state-of-the-art adaptive Kalman filter (AKF) technique is proposed to provide the axial distribution estimation of unmeasured system states in real-time, including delayed neutron precursor density, fuel and coolant temperatures, xenon and iodine concentrations, and reactivity, while utilizing only the available system measurements. On the other hand, this paper also further optimizes the mechanical shim (MSHIM) control strategy of the AP1000 by adopting an improved particle swarm optimization (IPSO) algorithm, which involves parameter optimizations for the first-order filters, lead-lag compensator, and differentiation-lag operator within the MSHIM framework. Simulation results indicate that (i) the optimized MSHIM control strategy using the IPSO outperforms the practically adopted approach, achieving significant improvements in control accuracy for load error and temperature error of the RC by 13.68 % and 21.31 %, respectively, while maintaining the same energy consumption, and (ii) under the proposed estimation algorithm in this paper, the estimates of system states provided by the AKF-based estimation algorithm exhibit strong agreement with their model-based values during the load following operation of the AP1000, with the maximum absolute relative error of 1.21 % merely, thereby verifying the proposed AKF-based estimation algorithm’s feasibility and accuracy.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128006"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144072345","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":"EL-NRF: Enhancing ensemble learning for regression with a noise reduction framework","authors":"Resul Özdemir ,&nbsp;Murat Taşyürek ,&nbsp;Veysel Aslantaş","doi":"10.1016/j.eswa.2025.128074","DOIUrl":"10.1016/j.eswa.2025.128074","url":null,"abstract":"<div><div>Ensemble learning aims to improve predictive accuracy by combining multiple models, with stacking being a widely adopted technique that employs a meta-learning framework. Despite significant advancements in stacking-based ensemble models, improving their robustness and generalization remains a persistent challenge. In this study, a two-phase noise reduction approach is proposed to improve the performance of stacking ensembles in regression tasks. In the first phase, feature-space noise is reduced through dimensionality reduction using Truncated Singular Value Decomposition (TSVD), which eliminates redundant and less informative components. In the second phase, sample-level noise is mitigated by applying a statistical thresholding method to identify and exclude high-residual instances. The proposed approach is evaluated on a real-world delivery time prediction dataset and six public benchmark datasets. Experimental results demonstrate that the integration of noise reduction techniques significantly enhances the predictive performance of stacking models, with improvements ranging from 1.65 % to 23.81 %, even in scenarios where conventional stacking fails to outperform its base learners. These results highlight the importance of noise reduction in improving the generalization capability of ensemble models, particularly in real-world regression problems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128074"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144069213","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":"Mask2Edge: Masking dependencies and dynamically capturing pixel differences in edge detection","authors":"Jianhang Zhou ,&nbsp;Hongwei Zhao ,&nbsp;Daikun Qu ,&nbsp;Long Xing","doi":"10.1016/j.eswa.2025.128041","DOIUrl":"10.1016/j.eswa.2025.128041","url":null,"abstract":"<div><div>Edge detection plays an important role in computer vision tasks. Deep learning-based edge detectors commonly rely on encoding the long and short-term dependencies of pixel values to mine contextual information. They strongly focus on all positions in the image, ignoring the potential issue of over-encoding. Furthermore, most of these models have not attempted to leverage the inherent properties of edges. In this paper, we introduce a query-based edge detector named Mask2Edge, which is capable of masking dependencies and dynamically capturing pixel differences. Specifically, we first devise a masking strategy based on the sparsity of edges to alleviate the over-encoding issue. We propose a Region-guided Masked Attention, which adapts to edge detection and is capable of constraining cross-attention with appropriate masking intensity to extract relatively complete local features. Subsequently, we design a structure to capture the pixel differences that can help identify edges. We introduce dynamic convolutions into edge detection and refine the application scope and generation method of attention weights to effectively perceive changes in pixel gradients. Extensive experiments demonstrate the superiority of Mask2Edge compared with state-of-the-art methods. The source code is available at https://github.com/zhoujh2020/Mask2Edge.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128041"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143948415","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":"SAAM-ReflectNet: Sign-aware attention-based multitasking framework for integrated traffic sign detection and retroreflectivity estimation","authors":"Joshua Kofi Asamoah ,&nbsp;Blessing Agyei Kyem ,&nbsp;Nathan David Obeng-Amoako ,&nbsp;Armstrong Aboah","doi":"10.1016/j.eswa.2025.128003","DOIUrl":"10.1016/j.eswa.2025.128003","url":null,"abstract":"<div><div>Traffic sign retroreflectivity is essential for roadway safety, particularly in low-light and adverse weather conditions. Traditional methods, such as handheld retroreflectometers and nighttime inspections, are labor-intensive, costly, and unsuitable for large-scale implementation. To address these limitations, we developed SAAM-ReflectNet, a deep learning framework that unifies traffic sign detection, classification, and retroreflectivity estimation into a single automated pipeline. Our RetroNet backbone, developed as part of this study, extracts robust spatial and semantic features to enhance feature representation. The Sign-Aware Attention Module we designed prioritizes critical traffic sign regions, improving detection and classification accuracy by focusing on the most relevant areas. Additionally, our multimodal fusion layers seamlessly integrate RGB imagery with LiDAR intensity data, enabling reliable retroreflectivity estimation. ReflectNet achieved a mean Average Precision (mAP) of 0.635 at IoU=0.5 and 0.522 across IoU thresholds from 0.5 to 0.95, alongside Root Mean Squared Errors (RMSE) of 0.169 for foreground and 0.147 for background reflectivity. Across 15 evaluation runs, performance improvements were statistically significant compared to all baselines (p &lt; 0.05), underscoring the consistency and reliability of ReflectNet.These findings underscore the reliability, scalability, and transferability of our approach, establishing ReflectNet as a transformative tool for intelligent transportation systems and proactive traffic sign maintenance.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128003"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068662","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":"Enhanced cloud security with Bi-Optimized Sand Cat Swarm and Conv-Bi-ALSTM deep learning models","authors":"Lubna Ansari","doi":"10.1016/j.eswa.2025.128128","DOIUrl":"10.1016/j.eswa.2025.128128","url":null,"abstract":"<div><div>As cyberattacks on cloud infrastructures become increasingly frequent and sophisticated, there is a growing demand for intelligent, scalable, and efficient intrusion detection systems (IDS). Traditional machine learning (ML) and deep learning (DL) models often struggle with computational complexity, data quality dependency, and scalability challenges. To address these limitations, this study introduces a novel AI-driven framework, Bi-Optimized SandCat-Conv-Bi-ALSTM (Bi-SC-CBALSTM), for enhanced threat detection in cloud environments. The framework begins with robust data preprocessing, employing Minkowski distance for redundancy elimination, nearest neighbor imputation for missing values, and min–max normalization for feature scaling. To resolve class imbalance, the ADASYN technique adaptively synthesizes minority samples near decision boundaries. For feature selection, the Binary Sand Cat Swarm Optimization (BOSCSA) algorithm efficiently extracts relevant features from high-dimensional data. These features are then passed into a hybrid deep model Conv-Bi-ALSTM, which combines convolutional layers for spatial feature extraction and a bidirectional LSTM enhanced with a 1 − tanh(x) function for improved sequential learning. Dropout layers are integrated to prevent overfitting, followed by a fully connected classifier. Experimental evaluation demonstrates that the proposed model achieves a balanced accuracy, precision, recall, and F1-score of 96 %, validating its robustness, scalability, and potential for real-time cloud threat detection.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128128"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068676","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":"Traffic forecasting with meta attentive graph convolutional recurrent network","authors":"Adnan Zeb ,&nbsp;Jianying Zheng ,&nbsp;Yongchao Ye ,&nbsp;Junde Chen ,&nbsp;Shiyao Zhang ,&nbsp;Xuetao Wei ,&nbsp;James Jianqiao Yu","doi":"10.1016/j.eswa.2025.128073","DOIUrl":"10.1016/j.eswa.2025.128073","url":null,"abstract":"<div><div>Traffic forecasting is essential for the development of intelligent transportation systems. However, existing forecasting models often struggle to effectively capture the complex spatial-temporal dependencies inherent in traffic data. Many current approaches are limited in their ability to model node-specific patterns and to simultaneously capture both short- and long-range dependencies. In this paper, we propose a novel traffic forecasting model, the Meta Attentive Graph Convolutional Recurrent Network (MAGCRN), which addresses these limitations through two key modules: (1) Node-Specific Meta Pattern Learning (NMPL) and (2) Node Attention Weight Generation (NAWG). The NMPL module captures the unique characteristics of each node in the traffic network by dynamically generating node-specific convolutional filters. The NAWG module enhances the model’s ability to capture both short- and long-range temporal dependencies by generating attention weights that connect node-specific features with those across the entire temporal dimension. Comprehensive experiments on six real-world traffic datasets demonstrate that MAGCRN consistently outperforms state-of-the-art baselines in both traffic flow and speed prediction tasks. The code is available at <span><span>https://github.com/Aazeb/MAGCRN</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"287 ","pages":"Article 128073"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084212","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":"Weakly supervised lesion localization and attribution for OCT images with a guided counterfactual explainer model","authors":"Limai Jiang ,&nbsp;Ruitao Xie ,&nbsp;Bokai Yang ,&nbsp;Juan He ,&nbsp;Huazhen Huang ,&nbsp;Yi Pan ,&nbsp;Yunpeng Cai","doi":"10.1016/j.eswa.2025.128129","DOIUrl":"10.1016/j.eswa.2025.128129","url":null,"abstract":"<div><div>Lesion localization plays an important role in computer aided diagnosis. Due to the lacking of lesion annotations, weakly supervised methods using only image-level annotations are demanded for a wide variety of applications, especially optical coherence tomography diagnosis. Most weakly-supervised methods rely on attribution analysis. However, current methods suffer from imprecise attributions and lead to poor localization quality. To address this problem, we introduce a lesion localization method based on a new explainable AI approach, termed Optical Coherence Tomography Class Association Embedding (OCT-CAE), leverages image-level annotations and a cycle generative adversarial network with subspace decomposition to fuse global knowledge and enable counterfactual generation. Each sample is encoded into a pair of subspaces where a low-dimensional common subspace is created to embed manifold of classification-related information and an individual subspace to embed individual-specific information. With the trained OCT-CAE, the codes in the two subspaces can be freely recombined to generate realistic images. These generated images retain the class-related features defined by the common subspace while preserving individual-specific characteristics from the individual subspace. Lesion localization is achieved by altering the common code to induce a class-flip in the generated images. By comparing the modified and original images, we can identify lesion regions without requiring regional annotations. Extensive experiments on two publicly available datasets demonstrate that OCT-CAE effectively disentangles latent information space in images, achieving state-of-the-art performance. Our code is available at <span><span>https://github.com/lcmmai/OCT-CAE</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"287 ","pages":"Article 128129"},"PeriodicalIF":7.5,"publicationDate":"2025-05-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144083673","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}