Expert Systems with Applications最新文献

{"title":"Towards a methodology for ethical artificial intelligence system development: A necessary trustworthiness taxonomy","authors":"Carlos Mario Braga ,&nbsp;Manuel A. Serrano ,&nbsp;Eduardo Fernández-Medina","doi":"10.1016/j.eswa.2025.128034","DOIUrl":"10.1016/j.eswa.2025.128034","url":null,"abstract":"<div><div>Recently, generative artificial intelligence (GenAI) has arisen and been rapidly adopted; due to its emergent abilities, there is a significantly increased need for risk management in the implementation of such systems. At the same time, many proposals for translating ethics into AI, as well as the first agreements by regulators governing the use of artificial intelligence (AI), have surfaced. This underscores the need for Trustworthy AI, which implies reliability, compliance, and ethics.</div><div>However, there is still a lack of unified criteria, and more critically, a lack of systematic methodologies for operationalizing trustworthiness within AI development processes. Trustworthiness is crucial, as it ensures that the system performs consistently under expected conditions while adhering to moral and legal standards. The problem of ensuring trustworthiness must be addressed as a preliminary step in creating a methodology for building AI systems with these desirable features. Based on a systematic literature review (SLR), we analyze the ethical, legal, and technological challenges that AI projects face, identifying key considerations and gaps in current approaches. This article presents a detailed and structured sociotechnical taxonomy related to the concept of Trustworthy AI, grounded in the analysis of all relevant texts on the topic, and designed to enable the systematic integration of ethical, legal, and technological principles into AI development processes. The taxonomy establishes a sociotechnical foundation that reflects the interconnected nature of technological, ethical, and legal considerations, and serves as the conceptual basis for CRISP-TAI, a proposed specialized development lifecycle currently under validation, aimed at systematically operationalizing trustworthiness principles across all phases of AI system engineering.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128034"},"PeriodicalIF":7.5,"publicationDate":"2025-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068661","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":"Multimodal learning analytics for students behavior prediction using multi-scale dilated deep temporal convolution network with improved chameleon Swarm algorithm","authors":"Thulasi Bharathi Sridharan ,&nbsp;Dr Periyaswamy Sappani Sharagaharajan Akilashri","doi":"10.1016/j.eswa.2025.128113","DOIUrl":"10.1016/j.eswa.2025.128113","url":null,"abstract":"<div><div>Despite the necessary knowledge, a variety of aspects affect the performance of the students on a task. Students’ on-task efforts have been noted to be significantly connected with their academic performance, demonstrating how effectively the students are involved in that task. Yet, these efforts given by students to perform a task cannot be seen directly. Multimodal knowledge may enable assessment of the student’s effort and offer extra perspectives into active learning. To provide new perspectives into the processes of learning premised on obtaining goals like behavioral trajectories, students’ achievement, support from teachers, feedback from students, involvement, and learning-task efficiency, a new deep learning-based multi-modal framework for data analytics of students’ behaviors are developed. At first, the multi-modal information is gathered from the relevant standard data sources, including video, sound, texts, EEG, eye movements, and facial information. For extracting the deep characteristics from the multi-modal input, the One Dimensional Convolutional Neural Network (1D-CNN) is utilized for feature extraction from audio, the Three Dimensional CNN (3D-CNN) is utilized for feature extraction from video, and Transformer-net is used for feature extraction from text. The proposed Modified Random Parameter-based Chameleon Swarm Algorithm (MRP-CSA) is used to optimize the weights before performing the weighted feature selection. The chosen characteristics are fed into the Adaptive Multi-scale Dilated Deep Temporal Convolution Network (AMDDTCN), which is utilized to identify the behavior of students while considering the expectations of the students, which also influences their involvement in the overall learning behavior evaluation stage. The implemented MRP-CSA is used to optimize the parameters inside the AMDDTCN during this evaluation phase. To ensure that the generated model is effective, the trial results are compared to the current multi-modal data analytics model. The developed model is evaluated using various performance metrics such as accuracy, precision, specificity, F1-score, precision, False Positive rate(FPR), False Negative Rate(FNR), Negative Predictive Value (NPV), False Discovery Rate (FDR) and Matthews Correlation Coefficient (MCC), and given the accuracy to be 97.2%. Thus, it is proved that the proposed model is promising over other traditional methods and has more abilities in identifying students’ behaviors by using a deep learning model with an optimization algorithm.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128113"},"PeriodicalIF":7.5,"publicationDate":"2025-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068645","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":"EPDPM-SinGAN: Enhancing urban street semantic segmentation with region-wise GANs feature","authors":"Khawaja Iftekhar Rashid,&nbsp;Chenhui Yang,&nbsp;Chenxi Huang","doi":"10.1016/j.eswa.2025.128053","DOIUrl":"10.1016/j.eswa.2025.128053","url":null,"abstract":"<div><div>Real-time semantic segmentation is essential in various applications, including autonomous driving and urban scene comprehension. This research focuses on the integration of discriminative and generative models to provide real-time semantic segmentation in complicated urban landscapes. This study introduces a novel model called EPDPM-SinGAN, which utilizes a SinGAN to extract context-aware features, together with an AdvVGG16 encoder and a U-Net decoder. The technique amplifies edge and texture characteristics to address typical challenges in semantic segmentation, particularly occlusions, and variations in object sizes. We incorporate Hierarchical Attention Mechanisms with Adaptive Feature Fusion to enhance the segmentation process and prioritize informative features. Finally, the Secondary Discriminative Pixel Mining (SDPM) module is introduced to target informative pixels for refined segmentation in complex urban scenarios. Our proposed technique EPDPM-SinGAN outperforms other segmentation models on the Cityscapes and CamVid datasets by achieving mIoU of 81.27 % and 78.7 % respectively, establishing itself as the current state-of-the-art.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"285 ","pages":"Article 128053"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143923244","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 multi-object soldier tracking algorithm based on trajectory association and improved YOLOv8n","authors":"Yu You,&nbsp;Jianzhong Wang,&nbsp;Shaobo Bian,&nbsp;Yong Sun,&nbsp;Zibo Yu,&nbsp;Weichao Wu","doi":"10.1016/j.eswa.2025.127877","DOIUrl":"10.1016/j.eswa.2025.127877","url":null,"abstract":"<div><div>In response to the challenges encountered in soldier tracking, including imprecise detection, high computational load, and frequent switching of soldier IDs, we propose a trajectory association and improved YOLOv8n-based soldier tracking algorithm, termed soldier tracking algorithm with YOLOv8-SD and HybridSORT-ST (STA-YH). The algorithm consists of two stages: soldier detection and soldier tracking. In the soldier detection stage, we propose an Efficient Dynamic C2f (ED-C2f) backbone network specifically designed to efficiently capture soldier features. Then, a novel Multi-branched Slim Context and Spatial Feature Calibration Network (MSCSFCN) is constructed to effectively fuse and align multi-scale soldier features. Furthermore, Group-Sparse Dynamic Head (GSDH) network is used to improve the attention of model to the soldier detection area. In the soldier tracking stage, we introduce the OSNet_IBN reidentification network and Adaptive Fading Kalman Filter (AFKF) algorithm into the HybridSORT and improve the state vector of the filter to reduce the frequency of ID switching for tracked soldiers. The results indicate that, in terms of soldier detection, compared with the baseline YOLOv8n, the improved YOLOv8-SD improved precision by 4.18% and mAP50-95 by 4.94% under the same computational load. This means that YOLOv8-SD is more accurate and has fewer missed or false detections. For soldier tracking, compared with the baseline, HybridSORT-ST demonstrates a 25.88% increase in HOTA, a 36.81% improvement in MOTA, and a 13.69% rise in IDF1, significantly improving the stability of continuous tracking of soldier in complex battlefield environments with dense movement and frequent occlusion, while meeting the requirements of lightweight design and tracking accuracy.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"285 ","pages":"Article 127877"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143942718","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 neural kernel framework for relation extraction","authors":"Kai Wang ,&nbsp;Yanping Chen ,&nbsp;Weizhe Yang ,&nbsp;Ruizhang Huang ,&nbsp;Yongbin Qin","doi":"10.1016/j.eswa.2025.128049","DOIUrl":"10.1016/j.eswa.2025.128049","url":null,"abstract":"<div><div>Relation extraction (RE) poses significant challenges due to the complexity of identifying semantic relationships between overlapping entity pairs within sentences. Traditional kernel-based methods effectively leverage instance-level similarities but heavily depend on manually designed kernel functions, requiring substantial domain expertise and limiting their flexibility. On the other hand, neural network approaches excel at automatically learning abstract feature representations but typically neglect critical instance-specific information, potentially missing valuable relational details during inference. These limitations motivate the exploration of a hybrid approach that effectively integrates the strengths of traditional kernels with the representational power of neural networks. In this paper, we propose a neural kernel framework designed to bridge this critical gap. The proposed framework employs neural kernels, whose parameters are initialized based on task-specific objectives and optimized through neural training procedures, enabling adaptive learning of similarity measures without relying on handcrafted kernel functions. By maintaining instance-level details through annotated training examples, neural kernels create discriminative, flexible decision boundaries tailored specifically to the relation extraction task. We further develop three complementary neural kernel components, instance kernel, description kernel, and cluster kernel, to show the advantages of kernel substitution. Extensive experiments on the ACE 2005, SemEval 2010, and CoNLL 2004 datasets demonstrate that our neural kernel framework significantly outperforms existing state-of-the-art methods, achieving F1-scores of 88.11 %, 91.08 %, and 98.63 %, respectively.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128049"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144069141","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":"Depolarizing power of anticonformity","authors":"Arkadiusz Lipiecki,&nbsp;Katarzyna Sznajd-Weron","doi":"10.1016/j.eswa.2025.127879","DOIUrl":"10.1016/j.eswa.2025.127879","url":null,"abstract":"<div><div>Political polarization hinders collective decision-making across multiple domains, from public health to environmental policy. Therefore, depolarization strategies are crucial and have been increasingly studied. Anticonformity, responding to social influence by opposing the opinions of others, has been associated with increased polarization, while its potential role as a depolarizing force has been largely overlooked. Although psychologists point to different forms of anticonformity, most computational models focus solely on xenophobia, prejudice against outsiders, which radicalizes opinions. Our work addresses this gap by considering another type of anticonformity – asserting uniqueness. We propose the counterintuitive hypothesis that increasing the disagreement by anticonforming to the influence group can reduce issue-based polarization. Within a family of computational models, we show that a depolarizing intervention based on promoting uniqueness may be more effective than traditional interventions, such as decreasing in-group favoritism or enhancing tolerance. We discuss the relevance of our findings through the lens of recent psychological experiments on strategic anticonformity, which demonstrate the potential of applying the proposed depolarizing intervention in real-world settings.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"285 ","pages":"Article 127879"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143942619","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":"FGDC: A fine-grained divide-and-conquer approach for extending NCO to solve large-scale Traveling Salesman Problem","authors":"Xinwei Chen ,&nbsp;Yurui Li ,&nbsp;Yifan Yang ,&nbsp;Li Zhang ,&nbsp;Shijian Li ,&nbsp;Gang Pan","doi":"10.1016/j.eswa.2025.127950","DOIUrl":"10.1016/j.eswa.2025.127950","url":null,"abstract":"<div><div>Large-scale Traveling Salesman Problem (TSP) applications are common and important in practice. Unfortunately, the time usage of the state-of-the-art heuristic solver LKH increases quadratically with the scale of the problem instance. Neural Combinatorial Optimization (NCO) has emerged as an efficient alternative to LKH, but it struggles to handle large-scale instances using zero-shot generalization. Divide-and-conquer is a classical paradigm for handling large-scale problems. However, integrating NCO with the divide-and-conquer paradigm is challenging. Fine-grained division is necessary to maintain the solution quality of the NCO solver on sub-problems. The K-Means algorithm is widely used for division as it delivers good results, but its time complexity grows to <span><math><mrow><mi>O</mi><mo>(</mo><msup><mi>N</mi><mn>2</mn></msup><mo>)</mo></mrow></math></span> in fine-grained division. Besides, merging strategies that rely on pre-defined rules miss opportunities to improve solution quality. In this paper, we present FGDC, a fine-grained divide-and-conquer approach for extending NCO to solve large-scale TSP. In the dividing procedure, we propose LocKMeans algorithm with <span><math><mrow><mi>O</mi><mo>(</mo><mi>N</mi><mo>)</mo></mrow></math></span> time complexity to construct small-scale sub-problems based on the density distribution of nodes. The solving procedure imposes minimal constraints on the NCO solver, which is used to solve these sub-problems with GPU parallel execution, allowing FGDC to serve as a plug-and-play tool for general NCO approaches. We propose an MST-based merging strategy which is enhanced from three different perspectives including merging combination, merging operator, and merging order. Experimental results demonstrate that FGDC outperforms existing methods in the fine-grained division scenario. Additionally, it is highly scalable over instances ranging from 1K to 1M nodes. When employing POMO (pre-trained on TSP-100) as the solver, FGDC surpasses the SOTA baseline H-TSP by a significant margin, yielding the best result for large-scale TSP within the NCO landscape.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 127950"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084412","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":"Arm rules and arm rule discovery in knowledge bases","authors":"Ya Chen ,&nbsp;Cunwang Zhang","doi":"10.1016/j.eswa.2025.128095","DOIUrl":"10.1016/j.eswa.2025.128095","url":null,"abstract":"<div><div>Rules are important for knowledge bases as they allow for logical reasoning and inference. Rule learning extracts patterns from existing knowledge, mainly focusing on first-order logic rules where atoms represent relations with variables for entity positions and rules are often in the form of closed paths. However, in reality, rules should not be limited to relations alone. They may also include specific entities and may take the form of an open path. In this study, we introduce a novel type of rule in knowledge bases, which we term “arm rules”. Specifically, an arm is the predicate-object portion of a knowledge triple, and an arm rule states that the existence of any arm following a head entity <span><math><mi>h</mi></math></span> forces or forbids the presence of certain other arms or relations following <span><math><mi>h</mi></math></span>. We also present the fundamental properties of these rules and introduce the rule discovery problem for such rules. Two algorithmic solutions are presented. One modifies the existing PCA confidence algorithm to this setting, while the other uses factor graph modelling framework and the belief propagation (or sum-product) algorithm therein. These two solutions are evaluated empirically using both real and synthetic data sets. Our experimental results demonstrate a superior performance of the factor graph approach.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"287 ","pages":"Article 128095"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084237","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":"KDCS-PPI: Knowledge distillation with counterfactual sampling for Protein-Protein Interaction prediction","authors":"Bin Deng ,&nbsp;Huifang Ma ,&nbsp;Ruijia Zhang ,&nbsp;Zhixin Li ,&nbsp;Liang Chang","doi":"10.1016/j.eswa.2025.127896","DOIUrl":"10.1016/j.eswa.2025.127896","url":null,"abstract":"<div><div>As the core of various biochemical reactions in life, Protein-Protein Interactions (PPIs) play a crucial role in maintaining the homeostasis of cellular functions, making the accurate prediction of PPIs particularly important. Traditional wet lab methods for predicting PPIs are time-consuming and costly. In contrast, PPI prediction methods utilizing Graph Neural Networks (GNNs) have exhibited promising performance and have increasingly emerged as the predominant approach in recent years. While GNNs rely on neighbor message aggregation, which can result in computational inefficiencies, Multilayer Perceptron (MLP) stand out for their time efficiency, as they do not require intricate handling of relational knowledge. However, MLPs often exhibit comparatively lower prediction accuracy. To leverage the advantages of both GNNs and MLPs in terms of effectiveness and efficiency, knowledge distillation techniques can be used to transfer the knowledge learned by GNNs to MLPs. During the knowledge distillation process, the knowledge transfer usually involves node feature embeddings rather than the interaction relationship knowledge between PPIs. Moreover, current methods frequently choose positive and negative samples for anchor nodes via random sampling, leading to suboptimal accuracy, especially for negative samples. To address this, we propose <em><strong>K</strong>nowledge <strong>D</strong>istillation with <strong>C</strong>ounterfactual <strong>S</strong>ampling for <strong>P</strong>rotein-<strong>P</strong>rotein <strong>I</strong>nteraction prediction</em> (KDCS-PPI). Our method facilitates the transfer of diverse relational knowledge between proteins during the knowledge distillation process and utilizes a counterfactual sampling strategy to select more pertinent positive and negative examples. Extensive experiments on three datasets demonstrate that KDCS-PPI can be applied to large-scale PPI prediction tasks and achieves significant improvements in both effectiveness and computational efficiency compared to other benchmark methods. Our source codes will be publicly available at <span><span>https://github.com/bin-db/KDCS-PPI</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"285 ","pages":"Article 127896"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143942646","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 novel 3D hyperchaotic map coupled with discrete tangent memristor: Dynamic analysis, DSP implementation, and image encryption application","authors":"Chenkai Zhang,&nbsp;Huibin Wang,&nbsp;Yiyan Zhang","doi":"10.1016/j.eswa.2025.128091","DOIUrl":"10.1016/j.eswa.2025.128091","url":null,"abstract":"<div><div>This study employs discrete memristors, derived from tangent functions, to generate a novel three-dimensional discrete hyperchaotic map. Initially, an analysis is conducted of the fixed-point attributes of the chaotic map. Subsequently, the dynamic behavior of the chaotic map is investigated, revealing phenomena such as hyperchaos, offset, state transitions, and the coexistence of attractors. The complexity and randomness of the chaotic sequence are evaluated, and the attractor trajectory is displayed on an oscilloscope via the DSP platform, thereby demonstrating the physical feasibility of the chaotic map. Consequently, an encryption algorithm for color images, incorporating selective diffusion processes, is developed. Lastly, a security analysis of the encryption algorithm is performed, affirming its robust security and resilience against various attack strategies.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"286 ","pages":"Article 128091"},"PeriodicalIF":7.5,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144068666","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}