Zheng Li , Lei Geng , Yanbei Liu , Feng Rong , Ming Ma , Jun Tong , Zhitao Xiao
{"title":"Uncertainty-guided denoising bi-classifier adversarial domain adaptation network for cross-domain fault diagnosis","authors":"Zheng Li , Lei Geng , Yanbei Liu , Feng Rong , Ming Ma , Jun Tong , Zhitao Xiao","doi":"10.1016/j.eswa.2025.129742","DOIUrl":"10.1016/j.eswa.2025.129742","url":null,"abstract":"<div><div>Intelligent fault diagnosis is crucial for ensuring the safety and reliability of modern industrial systems. However, the performance of deep learning models often significantly degrades due to the domain shift between training and testing data. Domain Adaptation (DA) methods, particularly bi-classifier adversarial networks, have proven effective in transferring knowledge from a labeled source domain to an unlabeled target domain. However, existing approaches often pay insufficient attention to target sample prediction accuracy, resulting in reduced feature discriminability and generalization. Additionally, due to the absence of labeled target data, most approaches rely on pseudo-labels, which are often noisy and unreliable, especially in the early stages of training. To address these issues, this paper proposes a novel uncertainty-guided denoising bi-classifier adversarial domain adaptation network (UGDBAN) for cross-domain fault diagnosis. Specifically, a feature generator based on Transformer layers is designed to capture long-range dependencies and local features. To mitigate the impact of noisy pseudo-labels, an uncertainty-based denoising pseudo-labeling mechanism is introduced to enhance the discriminability of features by redefining pseudo-labels and dynamically selecting high-confidence samples as clean samples. Building upon this denoised pseudo-label set, a Dirichlet uncertainty estimation-based class prototype alignment strategy is proposed to align domain features at the class level by selecting low-uncertainty samples representative of each class as prototypes. Extensive experiments demonstrate the effectiveness of UGDBAN, and comparative results with mainstream methods highlight its superiority.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129742"},"PeriodicalIF":7.5,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221481","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}
Patricia Rodriguez-Garcia , Angel A. Juan , Jon A. Martin , David Lopez-Lopez , Josep M. Marco
{"title":"AI-driven Optimization of project portfolios in corporate ecosystems with synergies and strategic factors","authors":"Patricia Rodriguez-Garcia , Angel A. Juan , Jon A. Martin , David Lopez-Lopez , Josep M. Marco","doi":"10.1016/j.eswa.2025.129593","DOIUrl":"10.1016/j.eswa.2025.129593","url":null,"abstract":"<div><div>This paper studies the optimization of project portfolios in corporate ecosystems by considering both strategic factors and return synergies between projects. We propose a hybrid method that combines machine learning with mathematical programming to address this enhanced form of project portfolio optimization. Unlike traditional approaches, which evaluate projects mainly based on individual risks and returns, our framework considers strategic priorities and the extra value created when projects reinforce each other. Machine learning models predict synergies, while exact optimization ensures consistent portfolio selection under resource and strategic constraints. A numerical proof-of-concept illustrates the methodology. Computational experiments show that portfolios designed with synergy and strategy in mind might achieve a significantly higher performance than portfolios that do not account for project synergies. The paper also examines computational efficiency and scalability, highlighting the approach’s potential for practical application in complex and dynamic corporate ecosystems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129593"},"PeriodicalIF":7.5,"publicationDate":"2025-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158442","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}
Yiran Liu , Dingshuo Liu , Mingrui Kong , Beibei Li , Qingling Duan
{"title":"Cross-scale content adaptive network for three-dimensional multi-object tracking and fish activity quantification","authors":"Yiran Liu , Dingshuo Liu , Mingrui Kong , Beibei Li , Qingling Duan","doi":"10.1016/j.eswa.2025.129774","DOIUrl":"10.1016/j.eswa.2025.129774","url":null,"abstract":"<div><div>Tracking and quantifying fish activity are vital for evaluating their health status and adaptability to the environment. However, most current research on fish tracking and activity quantification suffers from the limitation of being two-dimensional, losing crucial vertical or horizontal information. To facilitate tracking and quantitative analysis of fish activity in three-dimensional (3D) space, a cross-scale content-adaptive network-based 3D multi-object tracking method for fish is proposed, through which fish movements are quantified accordingly. Firstly, a cross-scale content-adaptive fusion network is proposed to accurately determine the fish positions from top-down and side views, thereby mitigating the issue of scale variation across different perspectives. Secondly, a hierarchical tracking method is implemented to obtain the 3D trajectories of the fish, addressing the challenge of cross-view identity matching. Finally, activity parameters in 3D space, including the activity quantity and trajectory length for individual fish, as well as the dispersion and cohesion for the fish group, are calculated. The proposed method was validated, achieving a Multi-Object Tracking Accuracy (MOTA) of 97.68% and an Identification F1 Score (IDF1) of 97.93%. For activity quantification, the Mean Absolute Error (MAE) was found to be 0.088 (unit weight·(cm/s)<sup>2</sup>), and the Root Mean Square Error (RMSE) was 0.1064 (unit weight·(cm/s)<sup>2</sup>). These results affirm the method’s adaption of fish features across scales for 3D tracking and activity analysis. With its efficient performance, our method presents as an instrument for activities such as fish behavior monitoring, selective breeding, and environmental assessment.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129774"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109327","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}
Jiangchuan Mei , Peizhong Yang , Hongmei Chen , Lizhen Wang
{"title":"Mining spatiotemporal dominant co-location patterns","authors":"Jiangchuan Mei , Peizhong Yang , Hongmei Chen , Lizhen Wang","doi":"10.1016/j.eswa.2025.129775","DOIUrl":"10.1016/j.eswa.2025.129775","url":null,"abstract":"<div><div>Spatial co-location pattern mining is an important branch of spatial data mining, which can identify spatial features that prevalently occur in proximity. Based on spatial co-location patterns, the research of dominant relationships mining within co-location patterns further considers the influence relationship among features. However, relying solely on spatial data to analyze the positions and distribution of features for mining dominant relationships is insufficient and may lead to incorrect patterns. To address this limitation, this paper introduces the temporal factor into the research of dominant relationships mining and proposes the spatiotemporal dominant co-location pattern mining (STDCPM). At first, we define the concepts of spatiotemporal dominant relationship from both temporal and spatial dimensions, and then propose the spatiotemporal dominant participation index to assess the prevalence of spatiotemporal dominant co-location patterns. Furthermore, we design two algorithms, the spatiotemporal dominant co-location pattern mining algorithm with level-by-level search and its improved version, i.e., the spatiotemporal dominant co-location pattern mining approach based on dual pruning and refining set (STDCPM-DPR), to ensure efficient mining in spatiotemporal datasets. The time complexity, correctness, and completeness of proposed algorithms are discussed. Extensive experiments on real-world datasets demonstrate the effectiveness of STDCPM and the efficiency of STDCPM-DPR algorithm.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129775"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109839","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}
Zhongyang Zhou , Bin Tang , Feiyu Chen , Wei Wang , Shangshang Zhao , Nanjun Yu
{"title":"scSCDT: Self-contrastive neural network with deep topology mining for scRNA-seq data clustering","authors":"Zhongyang Zhou , Bin Tang , Feiyu Chen , Wei Wang , Shangshang Zhao , Nanjun Yu","doi":"10.1016/j.eswa.2025.129751","DOIUrl":"10.1016/j.eswa.2025.129751","url":null,"abstract":"<div><div>Advancements in single-cell sequencing technologies have enabled researchers to better identify cells based on gene-level information. Cell clustering is a key task in single-cell analysis and plays an important role in distinguishing cell types. However, due to the high dimensionality and sparsity of scRNA-seq data, single-cell clustering remains a major challenge. Although many methods based on deep learning and machine learning have been developed for single-cell clustering, they often fail to capture the deep topological structure between cells, which limits clustering precision. In addition, most existing clustering approaches cannot effectively construct suitable sample pairs to optimize clustering models. To address these issues, we propose a topology-aware deep contrastive clustering model for single-cell data, named scSCDT. First, scSCDT employs a ZINB-based autoencoder to simultaneously learn cell embeddings and topological information, effectively handling the challenges posed by the high-dimensional and sparse nature of the data. Then, we introduce a dual clustering-guided loss to supervise the clustering task, combining a probabilistic soft assignment strategy and a hard pseudo-labeling strategy for optimization. Finally, based on the topological structure in the low-dimensional embedding space, we construct negative pairs within a single view and design a self-contrastive learning method to further improve clustering performance. We conduct extensive experiments on ten real scRNA-seq datasets and evaluate performance using four clustering metrics. The results indicate that scSCDT achieves strong clustering performance across multiple datasets, thereby facilitating more accurate cell type identification in single-cell transcriptomic analysis.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129751"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158647","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}
Yu-Qiang Wang , Yong-Ping Zhao , Tian-Ding Zhang , Yu-Wei Wang
{"title":"Time-balanced MSE for machinery imbalanced degradation trend prediction","authors":"Yu-Qiang Wang , Yong-Ping Zhao , Tian-Ding Zhang , Yu-Wei Wang","doi":"10.1016/j.eswa.2025.129783","DOIUrl":"10.1016/j.eswa.2025.129783","url":null,"abstract":"<div><div>Accurate degradation trend prediction (DTP) is crucial for optimizing equipment operation and maintenance. With the rapid development of artificial intelligence, many data-driven methods have been applied to machinery degradation trend prediction. In practice, most machineries are in the early stages of degradation, with only a few reaching the final stages, leading to a temporal imbalanced data distribution. Current research on imbalanced distributions mainly focuses on classification tasks. However, DTP involves multiple time-dependent continuous targets, making classification-based methods unsuitable. To address this issue, the degradation trend prediction task is reformulated as a multi-task problem and a novel time-balanced Mean Square Error (TBMSE) loss function is proposed. In each prediction task, the Gaussian Mixture Model (GMM) is used to fit the training label distribution. Additionally, the cumulative information noise for each prediction task is modeled using GMM, and an end-to-end network structure is designed to learn the GMM parameters. Experiments are conducted on the IMS bearing dataset and the turboprop engine dataset, demonstrating that the TBMSE loss effectively mitigates the issue of temporal imbalanced distribution in degradation trend prediction.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129783"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109326","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}
Chunbai Zhang , Haoyang Li , Chao Wang , Yang Zhou , Yan Peng
{"title":"FoRKER: Focused reasoner with knowledge editing and self-reflection","authors":"Chunbai Zhang , Haoyang Li , Chao Wang , Yang Zhou , Yan Peng","doi":"10.1016/j.eswa.2025.129771","DOIUrl":"10.1016/j.eswa.2025.129771","url":null,"abstract":"<div><div>Multi-hop question answering (MHQA) is a complex question answering (QA) benchmark that requires agents to integrate information from diverse sources and utilize cross-referencing reasoning to answer intricate questions. Existing MHQA-handling frameworks typically employ a <em>retrieve-read</em> paradigm. However, these efforts rooted in the retrieve-read paradigm are still constrained by: 1) <em>unstable document retrieval performance</em>, 2) <em>weak knowledge refinement capabilities</em>, and 3) <em>the absence of a reflection mechanism for error awareness</em>. To address these limitations, we propose <span>FoRKER</span> (<strong>Fo</strong>cused <strong>R</strong>easoner with <strong>K</strong>nowledge <strong>E</strong>diting and Self-<strong>R</strong>eflection), which is a plug-and-play framework. Specifically, we develop a novel progressive focusing mechanism to pinpoint highly relevant document resources and introduce knowledge editing techniques to further eliminate noise interference within textual information. Additionally, we design a novel prompting method, named Chain-of-Evidence (CoE), which is designed to augment the reasoning capabilities of <span>FoRKER</span>. Notably, the integration of Self-Reflection technology further endows <span>FoRKER</span> with the ability to learn and improve from its mistakes. Extensive experiments on widely-used datasets demonstrate that <span>FoRKER</span> achieves new state-of-the-art results in information retrieval and reading comprehension, while also exhibiting effective generalization. Exhilaratingly, on the MusiqueQA dataset, <span>FoRKER</span> demonstrates a 20 % improvement in Answering scores compared to the advanced competitors.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129771"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159096","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}
Jia Sheng Yang , Zihao Ning , Xu Xiao , Rui Zhong , Chenbo Xia , Ya Ding
{"title":"CG-TRAN: A novel multi-label retinal disease classification model with partially known pathologies","authors":"Jia Sheng Yang , Zihao Ning , Xu Xiao , Rui Zhong , Chenbo Xia , Ya Ding","doi":"10.1016/j.eswa.2025.129784","DOIUrl":"10.1016/j.eswa.2025.129784","url":null,"abstract":"<div><div>Early detection of retinal diseases is vital to preventing partial or permanent blindness. However, the diagnostic process is often impeded by the complexity of interrelated lesions and the challenge of incomplete or missing pathology labels, which require specialized expertise in ophthalmic diagnosis. To address these limitations, we propose CG-Tran, a novel multi-label classification model that leverages partially known pathology information to diagnose retinal diseases. This approach integrates a pathology graph neural network with graph-based feature extraction to handle partially known pathologies, enabling more accurate multi-label classification of retinal diseases. To model the intricate interrelationships among ocular diseases, CG-Tran employs BERT-GNN to learn label interactions and construct a comprehensive fundus pathology graph. Additionally, an enhanced attention mechanism incorporates known pathology label features, bridging the gap between incomplete pathology information and fundus image data. These innovations collectively empower the model to overcome the challenges of missing or incomplete pathology labels. The model’s performance is rigorously evaluated on the Multilabel Retinal Disease (MuReD) dataset. Results demonstrate that CG-Tran significantly improves diagnostic accuracy, especially as more pathology labels become available. Under conditions with 0% and 75% partially known labels, CG-Tran achieves mean average precision (mAP) scores of 69.9% and 72.1%, respectively—outperforming the baseline model by 1.0% and 1.9%. This innovative architecture excels in multi-label classification tasks, particularly in recognizing and distinguishing complex and interrelated retinal lesions with partially known pathology. It offers a promising solution for early detection and accurate diagnosis of retinal diseases, addressing critical limitations in existing diagnostic methods.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129784"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222087","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":"Spatiotemporal online fuzzy modeling with knowledge-driven differential evolution automatic clustering for distributed parameter systems","authors":"Gang Zhou, Xianxia Zhang, Bing Wang","doi":"10.1016/j.eswa.2025.129785","DOIUrl":"10.1016/j.eswa.2025.129785","url":null,"abstract":"<div><div>Distributed parameter systems are prevalent in various industrial processes and attract significant attention. However, these systems exhibit complex spatiotemporal coupling characteristics, and effectively determining the fuzzy rules of the antecedent set is crucial for improving modeling performance. Traditional clustering methods typically rely on empirical heuristics and are unable to adapt to dynamic system characteristics under changing environments. In high-dimensional and nonlinear scenarios, the number of fuzzy rule combinations grows exponentially, significantly increasing computational complexity. Therefore, an online spatiotemporal three-dimensional fuzzy modeling method based on knowledge-driven differential evolution automatic clustering and extreme learning machine (3D-OSADE-ELM) is proposed for the complex nonlinear distributed parameter system. First, an automatic clustering mechanism based on differential evolution and extreme learning machine initializes the fuzzy rules within the three-dimensional fuzzy system. Subsequently, a knowledge-driven archiving mechanism dynamically updates the fuzzy rules of the antecedent set during the online incremental learning phase. Finally, the spatial basis function is obtained by learning the output weight of the online extreme learning machine. The validation experiments conducted on the rapid thermal chemical vapor deposition reactor system and the nonisothermal packed-bed system demonstrate the effectiveness and superiority of the proposed method.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129785"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145118830","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}
Dawei Lin , Meng Yuan , Ying Chen , Xiaodong Zhu , Yuanning Liu
{"title":"Learning domain-invariant representation for generalizable iris segmentation","authors":"Dawei Lin , Meng Yuan , Ying Chen , Xiaodong Zhu , Yuanning Liu","doi":"10.1016/j.eswa.2025.129746","DOIUrl":"10.1016/j.eswa.2025.129746","url":null,"abstract":"<div><div>Cross-domain iris segmentation (CDIS) seeks to transfer knowledge from a labeled source dataset to an unlabeled target dataset. Existing CNN-based iris segmentation methods commonly assume that training and application stages share the same data distribution and modality setting, thus their performance may decline substantially on open-domain iris datasets unseen before. Furthermore, the process of annotating pixel-wise labels is labor-intensive and time-consuming, resulting in limited applicability of these methods in realistic scenarios. Therefore, we propose a generic domain adaptation iris segmentation framework (<em>DAIrisSeg</em>), which can be flexibly incorporated into existing methods. First, a domain-sensitive feature whitening strategy is proposed to effectively mitigate the domain-specific styles while preserving the domain-invariant content, thereby improving the model’s generalizability to unknown domain distribution. We then utilize the prototype estimation and the context-similarity learning adapter to produce reliable segmentation labels. In addition, DAIrisSeg incorporates prior constraints of the iris to further refine the segmentation results. Extensive experiments on three iris datasets demonstrate that the proposed method has shown consistent improvements over state-of-the-art (SOTA) methods.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129746"},"PeriodicalIF":7.5,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158641","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}