Expert Systems with Applications最新文献

{"title":"Solving the military medical evacuation dispatching, preemptive rerouting, redeploying, and delivering problem via tree-based machine learning and approximate dynamic programming approaches","authors":"Virbon B. Frial,&nbsp;Matthew J. Robbins,&nbsp;Phillip R. Jenkins","doi":"10.1016/j.eswa.2025.128582","DOIUrl":"10.1016/j.eswa.2025.128582","url":null,"abstract":"<div><div>Military medical evacuation (MEDEVAC) authorities face the challenge of efficiently dispatching aeromedical units and evacuating casualties to appropriate medical treatment facilities (MTFs). We examine a military MEDEVAC scenario wherein authorities must dispatch, preemptively reroute, and redeploy units while considering where to evacuate (or deliver) casualties, accounting for the capabilities and capacities of the MTFs (i.e., the military MEDEVAC DPR-D problem). To solve the problem efficiently, we formulate a discounted, infinite-horizon Markov decision process (MDP) model and employ approximate dynamic programming (ADP) solution techniques that integrate tree-based value function approximation schemes within an approximate policy iteration (API) framework: Random Forest (API-RF) and Extreme Gradient Boosting (API-XGB). Using domain knowledge-based basis functions, we enhance the explainability of these approximation schemes. We construct a representative scenario of high-intensity operations in Bosnia-Herzegovina to demonstrate the applicability of our MDP model and compare the efficacies of our ADP solution techniques. The results show that API-RF and API-XGB significantly outperform the current benchmark myopic policy (i.e., assign the closest unit and MTF to the casualty location) across all 36 problem instances. Moreover, API-XGB consistently outperforms API-RF in 32 instances, achieving statistical significance at the 95 % confidence level for 21 of them. The explainability of these tree-based schemes highlights key features that influence DPR-D policies, such as the casualty queue length and the number of available MTF beds, whose importance shifts depending on the casualty arrival intensity. Our research offers valuable insights and potential modifications for future military MEDEVAC operations.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"292 ","pages":"Article 128582"},"PeriodicalIF":7.5,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144322785","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":"Enhancing mammogram classification using explainable Conditional Self-Attention Generative Adversarial Network","authors":"K.K. Sreekala,&nbsp;Jayakrushna Sahoo","doi":"10.1016/j.eswa.2025.128640","DOIUrl":"10.1016/j.eswa.2025.128640","url":null,"abstract":"<div><div>Globally, breast cancer is one of the leading causes of death in women. Thus, there is an urgent requirement for precise and comprehensible diagnostic instruments. This paper introduces a new deep learning model, an Explainable Conditional Self-Attention Generative Adversarial Network (ExCSA-GAN), suggested for the classification of breast cancer using mammography images. The used input mammograms were drawn from the publicly available CBIS-DDSM breast cancer image dataset and the LHD dataset. Noise in the images is minimized with Window-Aware Guided Bilateral Filtering (WAGBF). These images are then further segmented for cancerous regions through the use of the Median-Average 2D Otsu’s Otsu-based segmentation (MA-2D-O). Finally, classification is done using ExCSA-GAN, which performs well on the target classification metric while being interpretable. The model hyperparameters are fine-tuned using the Greylag Goose Optimization (GGO) algorithm, which leads to optimal performance. To enhance the transparency of predictions, the proposed approach integrates four explainable algorithms: Gradient-Weighted Class Activation Mapping (Grad-CAM), Shapley Additive Explanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), and Layer-Wise Relevance Propagation (LRP). Comparing ExCSA-GAN to traditional deep learning models, experimental findings show that it improves accuracy by 9.8% and reduces false negative rate (FNR) by 12.5%. The superiority of the proposed approach is validated by experimental results using the core metrics, which include accuracy, Matthews Correlation Coefficient (MCC), precision, sensitivity, specificity, F-measure, computational complexity, and computation time. This approach offers better performance and improved interpretability for clinical applications.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"293 ","pages":"Article 128640"},"PeriodicalIF":7.5,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144338263","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 pre-trained data deduplication model based on active learning","authors":"Haochen Shi ,&nbsp;Xinyao Liu ,&nbsp;Fengmao Lv ,&nbsp;Hongtao Xue ,&nbsp;Jie Hu ,&nbsp;Shengdong Du ,&nbsp;Tianrui Li","doi":"10.1016/j.eswa.2025.128628","DOIUrl":"10.1016/j.eswa.2025.128628","url":null,"abstract":"<div><div>In the era of big data, the issue of data quality has become increasingly prominent. One of the main challenges is the problem of duplicate data, which can arise from repeated entry or the merging of multiple data sources. These ”dirty data” problems can significantly limit the effective application of big data. To address the issue of data deduplication, we propose a pre-trained deduplication model based on active learning, which is the first work that utilizes active learning to address the problem of deduplication at the semantic level. The model is built on a pre-trained Transformer and fine-tuned to solve the deduplication problem as a sequence to classification task, which firstly integrate the transformer with active learning into an end-to-end architecture to select the most valuable data for deduplication model training, and also firstly employ the R-Drop method to perform data augmentation on each round of labeled data, which can reduce the cost of manual labeling and improve the model´s performance Experimental results demonstrate that our proposed model outperforms previous state-of-the-art (SOTA) for deduplicated data identification, achieving up to a 28 % improvement in Recall score on benchmark datasets.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"292 ","pages":"Article 128628"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144314181","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":"Prior-assisted unpaired image dehazing framework for enhanced visibility in real-world hazy scenarios","authors":"Pengyang Ling ,&nbsp;Haoxuan Wang ,&nbsp;Huaian Chen,&nbsp;Yuxuan Gu,&nbsp;Yi Jin,&nbsp;Jinjin Zheng","doi":"10.1016/j.eswa.2025.128488","DOIUrl":"10.1016/j.eswa.2025.128488","url":null,"abstract":"<div><div>To facilitate a stable dehazing performance in real scenarios, this article proposes a novel prior-assisted unpaired image dehazing framework (PAUD), which obtains superior dehazing performance directly from real unpaired hazy/clear images. Specifically, a fast haze modulation (FHM) scheme is presented, which enables fast and flexible modulation in haze concentration for effortless production of diverse hazy samples, promoting the capability in dealing with complex scenarios. Moreover, an adaptive prior matching (APM) mechanism has been developed to alleviate the risk of misguidance caused by prior failure. This mechanism performs soft constraint with prior-based transmission by estimating a pixel-wise credibility map. Extensive experiments demonstrate that the proposed method outperforms start-of-the-art methods in achieving enhanced visibility while requiring fewer parameters, providing effective and efficient visibility improvement under various hazy conditions.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"291 ","pages":"Article 128488"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291155","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":"Human-informed skill discovery: Controlled diversity with preference in reinforcement learning","authors":"Maxence Hussonnois ,&nbsp;Thommen George Karimpanal ,&nbsp;Mayank Shekhar Jha ,&nbsp;Santu Rana","doi":"10.1016/j.eswa.2025.128604","DOIUrl":"10.1016/j.eswa.2025.128604","url":null,"abstract":"<div><div>Autonomously learning diverse behaviours without an extrinsic reward signal has been a problem of interest in reinforcement learning. However, the nature of learning in such mechanisms is unconstrained, often resulting in the accumulation of several unusable, unsafe or misaligned skills. In order to avoid such issues and to ensure the discovery of safe and human-aligned skills, it is necessary to incorporate humans into the unsupervised training process, which remains a largely unexplored topic. In this work, we propose Controlled Diversity with Preference (CDP)<span><math><msup><mrow></mrow><mrow><mn>1</mn><mo>,</mo><mn>2</mn></mrow></msup></math></span>, a novel, collaborative human-guided mechanism for an agent to learn a set of skills that is diverse as well as desirable. The key principle is to restrict the discovery of skills to regions that are deemed to be desirable as per a preference model trained using human preference labels on trajectory pairs. We evaluate our approach on 2D navigation and Mujoco environments and demonstrate the ability to discover diverse, yet desirable skills. We also provide principled guidelines for selecting suitable hyperparameter values along with comprehensive sensitivity analyses of the various factors influencing the performance of our approach.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"292 ","pages":"Article 128604"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313991","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":"Deep learning techniques for DDoS attack detection: Concepts, analyses, challenges, and future directions","authors":"Xingbing Fu ,&nbsp;Supeng Lou ,&nbsp;Jiaming Zheng ,&nbsp;Cheng Chi ,&nbsp;Jie Yang ,&nbsp;Dong Wang ,&nbsp;Chenming Zhu ,&nbsp;Butian Huang ,&nbsp;Xiatian Zhu","doi":"10.1016/j.eswa.2025.128469","DOIUrl":"10.1016/j.eswa.2025.128469","url":null,"abstract":"<div><div>DDoS (Distributed Denial of Service) attacks are increasingly becoming a major threat in the field of cybersecurity. They overwhelm target servers by sending large-scale requests from multiple locations, causing the servers to become unresponsive. The distributed nature of DDoS attacks also makes detection and defense even more challenging. As the damage caused by such attacks grows, the development of efficient detection and mitigation mechanisms has become an urgent priority. While traditional machine learning methods are useful, they still require manual feature extraction, which not only involves significant human intervention, but also is time-consuming. In contrast, deep learning offers an automated approach to feature extraction and can learn more abstract patterns, which leads to improved detection performance. Therefore, this paper reviews and analyzes existing deep learning methods for DDoS attack detection. We provide a comprehensive analysis of the various types of DDoS attacks and explore different deep learning models employed for attack detection. Additionally, we explore techniques such as federated learning that can be integrated with deep learning, and analyze their related literature in DDoS attack detection. Finally, we specify future research directions on DDoS attack detection using deep learning.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"291 ","pages":"Article 128469"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291156","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":"Reinforcement learning based early classification framework for power transformer differential protection","authors":"Xiaopeng Wang ,&nbsp;Anyang He ,&nbsp;Zongbo Li ,&nbsp;Zaibin Jiao ,&nbsp;Na Lu","doi":"10.1016/j.eswa.2025.128632","DOIUrl":"10.1016/j.eswa.2025.128632","url":null,"abstract":"<div><div>The balance between response speed and diagnosis accuracy forms a critical concern in transformer protection. However, prevailing AI-based transformer protection methods tend to adopt fixed data length to extract electrical quantity information, thus impeding prompt responsiveness to situations where discriminative fault features emerge in the early stages. This study formulates transformer protection as a Markov decision process and proposes an Early Classification Proximal Policy Optimization (ECPPO) framework to utilize reinforcement learning (RL) for data-length adaptive transformer protection with timely action and notable high accuracy. However, the limited generalization of RL algorithms poses a significant issue in the transformer protection scenario. While enhancing the feature extraction capability of a model is essential for improving its generalization ability, ECPPO constructs a two-stage training paradigm to augment the policy model accordingly. In the first stage, a multi-task deep learning framework trains a feature-extraction module with normalization layers employing fault label information and a signal reconstruction task to enrich the feature representation. In the second stage, the pre-trained feature-extraction module is transferred to the agent model with frozen weights, and PPO training is performed. Additionally, to improve the utilization efficiency of samples, a period-circle-shift data augmentation method is proposed, which enhances the generalization capability by cyclically reconstructing data in periodic sequences. To validate the proposed framework, a series of experiments were conducted using simulation data generated by PSCAD/EMTDC software as the training data and practical data generated by experimental transformer system as the testing data. The experimental results demonstrate a significantly enhanced testing accuracy of 99.19 %, coupled with an average response time of 12.10 ms, indicating that the ECPPO algorithm not only achieves superior accuracy but also effectively reduces the average response time. Furthermore, the results highlight its robust generalization capability when transitioning from simulation to experimental systems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"292 ","pages":"Article 128632"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144331233","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":"Maximum relevant diversity aware, multi-video summarization using clustering and evolutionally multi-objective optimization","authors":"Aziz M. Qaroush,&nbsp;Mohammad Jubran,&nbsp;Qutaiba Olayyan,&nbsp;Osama Qutait","doi":"10.1016/j.eswa.2025.128631","DOIUrl":"10.1016/j.eswa.2025.128631","url":null,"abstract":"<div><div>With the exponential increase in video data, efficient video summarization techniques are crucial for handling and analyzing large-scale collections. Multi-video summarization poses unique challenges due to inter-video correlations, redundancy, and variability in content. This paper introduces a novel multi-video summarization approach that combines clustering with evolutionary multi-objective optimization to produce concise, diverse, and informative summaries. Our method begins by segmenting videos into scenes and representing each using I3D CNN features for spatial and temporal dynamics. Through clustering, similar segments across videos are grouped to maximize content diversity while minimizing redundancy. Key video summarization features-visual attention, coverage, and diversity-are then extracted from clusters and segments, and balanced using a multi-objective optimization algorithm. The final summary is generated by selecting a subset of optimized solutions based on a predefined summary ratio, ensuring relevance, coverage, and diversity. Extensive experiments on the Tour20 dataset demonstrate the superiority of our approach, highlighting its potential to address the complexities of multi-video summarization.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"292 ","pages":"Article 128631"},"PeriodicalIF":7.5,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313923","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":"USPDB: A novel U-shaped equivariant graph neural network with subgraph sampling for protein-DNA binding site prediction","authors":"Jia Mi ,&nbsp;Chang Li ,&nbsp;Han Wang ,&nbsp;Ying Du ,&nbsp;Chong Chu ,&nbsp;Jing Wan ,&nbsp;Kunfeng Wang","doi":"10.1016/j.eswa.2025.128554","DOIUrl":"10.1016/j.eswa.2025.128554","url":null,"abstract":"<div><div>Protein-DNA binding directly influences the normal functioning of biological processes by regulating gene expression. Accurate identification of binding sites can reveal the mechanisms of protein-DNA interactions and provide a clear direction for drug target development. However, traditional experimental methods are time-consuming and costly, necessitating the development of efficient computational methods. Although existing computational methods have made significant progress in the field of protein binding site prediction, they have difficulty extracting key residue features and atomic-level features. To address this, we propose a novel method, USPDB, based on a U-shaped Equivariant Graph Neural Network(U-EGNNet) and Subgraph Sampling for Protein-DNA Binding Site Prediction. USPDB reformulates the binding site prediction task by converting the protein into a graph and performing a binary classification for each residue. It leverages protein large language models, such as Protrans, ESM2, and ESM3, to extract sequence and structural features. The General Equivariant Transformer (GET) module is employed to capture geometric features of residues and atoms. Additionally, the U-EGNNet, composed of EGNN and Subgraph Sampling, is utilized to preserve more global information while sampling subgraphs that contain key residues for further computation. Experimental results on DNA_test_181 and DNA_test_129 datasets demonstrate that USPDB achieves prediction accuracies of 0.532 and 0.361, respectively, outperforming all baseline methods. Through interpretability analysis, we observed that USPDB effectively focuses on residues within DNA-binding domains without requiring prior knowledge, thereby enhancing the performance of DNA-binding protein prediction. The code is publicly available at the following link: <span><span>https://github.com/MiJia-ID/USPDB</span><svg><path></path></svg></span></div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"291 ","pages":"Article 128554"},"PeriodicalIF":7.5,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144280412","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":"Bilevel reinforcement learning imaging method for electrical capacitance tomography","authors":"Jing Lei ,&nbsp;Qibin Liu","doi":"10.1016/j.eswa.2025.128614","DOIUrl":"10.1016/j.eswa.2025.128614","url":null,"abstract":"<div><div>Despite demonstrating considerable promise as a tomography technology for multiphase flow parameter measurements, electrical capacitance tomography is constrained by the inherent suboptimal image reconstruction quality. In order to fully harness its potential, the image reconstruction problem is modeled as a new bilevel fractional optimization problem. This new model integrates the advantages of bilevel optimization and fractional optimization, takes into account the inaccuracy of the measurement model and measurement data, fuses measurement principles with deep learning, learns the model parameters adaptively, achieves the multi-source information fusion, mitigates the ill-posed property of the image reconstruction problem, improves the automation and robustness of the model, and enhances the model’s ability to handle complex measurement scenarios. In order to augment image priors and improve the comprehensive reconstruction performance, based on the regularization by denoising, deep convolutional neural network, as an efficient denoiser, is integrated into the reconstruction model. Following the algorithm unfolding principle, we convert the proposed bilevel fractional optimization problem into a single-level nonlinear optimization problem, which effectively handles the nested structure between the upper and lower level optimization problems and reduces the computational complexity. A new optimizer is proposed to solve this transformed optimization problem. It integrates reinforcement learning and differential evolution algorithm, and is able to adaptively adjust the algorithm parameters by leveraging the interaction and feedback between the parameter configures and the computational results, thus improving the performance of the algorithm and the quality of the optimal solution. Empirical evaluations demonstrate that our novel approach not only yields enhanced imaging quality but also exhibits superior noise resilience when benchmarked against widely-adopted imaging algorithms, while maintaining consistent performance across various scenarios. Our study offers a holistic solution for improving the overall efficacy of image reconstruction tasks by the synergistic fusion of supervised learning methodologies and optimization principles. This fusion not only maximizes the capabilities of the advanced measurement technology but also unlocks its full potential for achieving high-quality reconstruction results.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"293 ","pages":"Article 128614"},"PeriodicalIF":7.5,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144338260","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}