Expert Systems with Applications最新文献

{"title":"Multi-graph fusion guided robust adaptive learning for subspace clustering","authors":"Jianyu Miao ,&nbsp;Xiaochan Zhang ,&nbsp;Chao Fan ,&nbsp;Tiejun Yang ,&nbsp;Yingjie Tian ,&nbsp;Yong Shi ,&nbsp;Mingliang Xu","doi":"10.1016/j.eswa.2025.129918","DOIUrl":"10.1016/j.eswa.2025.129918","url":null,"abstract":"<div><div>Subspace clustering is an advanced technique that identifies clusters embedded within a union of low-dimensional subspaces of the original data space, thereby revealing its intrinsic structure. Spectral clustering-based methods have gained significant attention in computer vision, image processing and pattern recognition due to their promising performance. However, existing approaches, which typically rely on self-representation for representation coefficient learning, often lack robustness and struggle to comprehensively characterize complex data structures. Traditional reconstruction loss based on the Frobenius or <span><math><msub><mi>ℓ</mi><mn>1</mn></msub></math></span> norm are susceptible to noise and outliers. Furthermore, many methods underutilize inherent data characteristics for capturing local geometric structures and adapting to intricate data relationships. To address these limitations, this paper proposes a novel subspace clustering approach, named Multi-graph Fusion Guided Robust Adaptive Learning (MFGRAL), which integrates robust adaptive representation and multi-graph fusion within a unified framework. Specifically, a non-convex logarithmic loss function is adopted to enhance robustness against noise and outliers. To better preserve local manifold structures, a multi-graph fusion strategy is developed to guide the adaptive graph learning process. This facilitates the learning of more discriminative low-dimensional embeddings and enhances the capacity to capture complex neighborhood relationships. An effective and efficient optimization algorithm based on Alternating Direction Method of Multipliers (ADMM) is developed to solve the proposed model. Extensive experimental results on several benchmark datasets demonstrate the effectiveness of the proposed MFGRAL and its superiority over state-of-the-art methods.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129918"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145268337","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":"PDHG: An Ethereum phishing detection approach via heterogeneous graph transformer","authors":"Lei Wang,&nbsp;Yihan Mi,&nbsp;Yanan Zhang,&nbsp;Jialin Zhang","doi":"10.1016/j.eswa.2025.129919","DOIUrl":"10.1016/j.eswa.2025.129919","url":null,"abstract":"<div><div>Phishing scams have emerged as a significant threat within the Ethereum ecosystem. Cutting-edge Ethereum phishing scams detection techniques mostly treat accounts in Ethereum as homogeneous nodes in transaction graphs. Existing detection approaches model Ethereum transaction records as homogeneous transaction graphs and use graph representation learning for account classification. However, those approaches often overlook the heterogeneity between accounts and transactions, making it difficult to capture the diversity of interactions and features among accounts. In this paper, a heterogeneous graph transformer (HGT)-based phishing account identification approach called PDHG is proposed. Specifically, PDHG models the transaction network between accounts as a heterogeneous graph based on different attributes of Ethereum accounts, allowing for a more comprehensive description of the structure and behavioral patterns of the transaction network. To enhance the explainability, PDHG leverages PDHGexplainer as the explainer for the detection results. We compare PDHG with other existing detection models. The experimental results demonstrate that PDHG achieves an AUC score of 96.04 % and a recall score of 89.87 %, surpassing the state-of-the-art approaches.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129919"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145268425","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":"EEG-based emotion identification from nerve conduction mechanisms: A gustatory-emotion coupling model combined with multiblock attention module","authors":"Wenbo Zheng ,&nbsp;Yong Peng ,&nbsp;Ancai Zhang ,&nbsp;Quan Yuan","doi":"10.1016/j.eswa.2025.129855","DOIUrl":"10.1016/j.eswa.2025.129855","url":null,"abstract":"<div><div>Electroencephalogram (EEG)-based emotion identification enables accurate emotional interaction in brain-computer fusion by decoding brain signals, thereby enhancing the intelligence of human-computer collaboration. Data augmentation (DA) techniques offer a promising solution to the challenge of data scarcity in emotion identification. However, traditional DA methods often overlook the physiological mechanisms underlying EEG data, limiting their effectiveness and constraining the performance of emotion classification. To address this, a DA model based on human nerve conduction mechanisms (NCMs), named the gustatory-emotion coupling model and multiblock attention module (GECM-MBAM), is proposed to improve the performance of emotion identification. First, the 1/<em>f</em> characteristics and synchronization of brain responses are reproduced in the GECM output when stimulated by EEG. The bionic performance of the model in EEG processing is validated, demonstrating brain-like perception of EEG signals via the GECM. Second, the MBAM is designed based on the characteristics of the GECM output, facilitating data augmentation of emotion-related EEG. Comparative experiments demonstrate that GECM-MBAM remarkably outperforms multiple existing DA models in recognition accuracy (<em>p</em> &lt; 0.05), confirming its effectiveness and superiority in EEG data augmentation. Finally, when compared with state-of-the-art algorithms and in ablation studies, GECM-MBAM demonstrates superior performance in emotion recognition. Specifically, GECM-MBAM attains accuracies of 96.91 % and 94.52 %, recalls of 96.23 % and 93.86 %, and kappa coefficients of 95.45 % and 94.29 % on the SEED and SEED-IV datasets, respectively. In conclusion, the performance of emotion identification is improved using the GECM-MBAM, offering a novel bionic processing approach for affective computing.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129855"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222085","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":"FSTrack: Visual tracking with feature fusion and adaptive selection","authors":"Jian Shi ,&nbsp;Yang Yu ,&nbsp;Bin Hui ,&nbsp;Junze Shi ,&nbsp;Haibo Luo","doi":"10.1016/j.eswa.2025.129895","DOIUrl":"10.1016/j.eswa.2025.129895","url":null,"abstract":"<div><div>Visual object tracking represents a critical research domain within computer vision, with significant applications spanning security surveillance, autonomous navigation, and other fields. Throughout the tracking process, distractors and target appearance variations frequently arise, rendering sole reliance on initial templates unreliable. Therefore, the effective integration of spatiotemporal information and search region features plays a crucial role in achieving robust long-term single-object tracking. However, most existing methods indiscriminately incorporate all historical features as spatiotemporal context, potentially introducing irrelevant or redundant information that undermines tracking reliability. To address this limitation while more effectively exploiting backbone features, we propose FSTrack, which leverages feature fusion to enhance search features and adaptively selects features to strengthen spatiotemporal features. First, we integrate multi-level backbone features through feature fusion and enhance feature resolution, thereby fully exploiting the multi-scale features of the backbone networks. Second, we introduce an adaptive feature selection mechanism that dynamically identifies and emphasizes discriminative historical features, enhancing the robustness of spatiotemporal modeling under diverse tracking scenarios. Third, we propose a globally contextual prediction head that overcomes the limitation of the limited receptive field inherent in conventional CNN-based heads and further improving the overall performance. Extensive experiments demonstrate the superiority of FSTrack. On mainstream benchmark datasets such as GOT-10k, TrackingNet, and LaSOT, our approach outperforms mainstream models using both the same and higher resolution inputs in terms of speed and accuracy, achieving state-of-the-art results on tracking benchmarks.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129895"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267964","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":"Transfer learning based standard-essential patent prediction with prior transfer direction learning","authors":"Weidong Liu ,&nbsp;Xiaoyu Fan ,&nbsp;Kai Wang ,&nbsp;Hongjun Sun ,&nbsp;Keqin Gan ,&nbsp;Cuicui Jiang ,&nbsp;Fangyuan Lei","doi":"10.1016/j.eswa.2025.129869","DOIUrl":"10.1016/j.eswa.2025.129869","url":null,"abstract":"<div><div>Standard-Essential Patent Prediction (SEPP) holds strategic significance for technological development and international market competition. Traditional SEPP models learned from Standard-Essential Patents (SEPs) with country-specific distribution differences result in different prediction accuracy. Therefore, we propose two propositions: (1) Can transfer learning be leveraged to improve prediction performance of lower-accuracy countries. (2) Can different transfer directions achieve different transfer learning performances. To address these, we propose a transfer learning based SEPP ewith prior transfer direction learning (TLSEPP-PTDL) model. The model uses a mixed transfer learning method, achieving an average accuracy of 92.03 % on four datasets, surpassing the state-of-the-art (SOTA) by 2.03 % and improving precision, recall, and F1-score by 4.25 %, 0.33 %, and 2.25 %, respectively. Moreover, we conduct experiments across countries with different patent volume, standardization rate, and standardization speed, resulting in positive transfer when transfer learning uses source domains with (1) high volume, high rate, and high speed; (2) high volume, low rate, and high speed; (3) low volume, high rate, and high speed.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129869"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145268543","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 belief rule-based system for online and centralized collaborative performance assessment of networked physical systems subject to nonideal channels","authors":"Haoran Zhang ,&nbsp;Lining Xing ,&nbsp;Jian Wu ,&nbsp;Ruohan Yang ,&nbsp;Zhichao Feng","doi":"10.1016/j.eswa.2025.129803","DOIUrl":"10.1016/j.eswa.2025.129803","url":null,"abstract":"<div><div>Networked physical systems (NPSs) are widely applied in modern engineering practices characterized by intensive domain knowledge and imperfect observational data. Meanwhile, collaborative performance assessment provides strong support for them to operate safely and stably over a long period of time. For a specific NPS and its corresponding online and centralized collaborative performance assessment system, the existence of interference and noise in the real-world channel that is rather nonideal inevitably obstructs the smooth progress of the assessment. To this end, in this paper, a symbolic systematic solution is proposed resorting to an improved version of the belief rule base with continuous inputs (BRB-CI). First, the extrapolation module is enhanced by integrating a matched filtering-based link. Second, the existing robustness analysis for systems based on the fundamental belief rule base is extended to systems based on the BRB-CI. Third, the optimization module is ameliorated by designing a multimetric-balanced pattern of the grey wolf optimizer with interpretability reinforcement. Ultimately, by choosing an instance of NPSs in the field of aerospace with continuous time dynamics, pertinent empirical studies are carried out to substantiate the good engineering practicability of our proposal. Note that this paper is the first piece inquiring into belief rule-based systems such a class of expert systems for online and centralized cooperative performance assessment of NPSs with continuous time dynamics such an application, with considerable attention paid to the nonideality of real-world channels.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129803"},"PeriodicalIF":7.5,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222084","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":"The knowledge-driven adaptive late acceptance iterative hill-climbing heuristics for the bus and ADR collaborative delivery problem","authors":"Lijun Pan ,&nbsp;Changshi Liu ,&nbsp;Yifan Zhang ,&nbsp;Shun Li","doi":"10.1016/j.eswa.2025.129810","DOIUrl":"10.1016/j.eswa.2025.129810","url":null,"abstract":"<div><div>Urban-rural bus transit services encounter a dilemma between the necessity for enhanced services and the challenge of low profitability due to scant travel demand. Combining freight transportation with passenger services can enhance the efficiency and profitability of buses in both urban and rural areas, while also reducing environmental impacts. A case in point is the integration of freight deliveries into rural bus networks in China. Concurrently, with the advancement of autonomous delivery robot (ADR) technology, there is a growing deployment of ADRs for last-mile delivery purposes. In this paper, we have studied a new collaborative passenger and freight transportation problem involving buses and ADRs, namely, the bus and ADR collaborative delivery problem (BACDP). In this scenario, a bus route transports several ADRs, which carry multiple parcels, to distribution regions for door-to-door delivery, each ADR boards a bus to reach the sub-region and then boards another bus to return to the distribution center. We have proposed a mathematical model for BACDP, which can be decomposed into a master problem and a sub-problem. and the condition that the optimal solution to the master problem is also the optimal solution to the original problem has been proved. To tackle the BACDP effectively, we designed a novel three-stage iterative method, guided by adaptive late acceptance hill-climbing heuristics (ALAHH). Specifically, at the first stage, the k-means++ and Hamiltonian graph-guided algorithms are used to cluster customers; at the second stage, the variable neighborhood search plans the ADRs’ routes; at the third stage, we utilize the solver to address subproblems, and the evaluation and invocation mechanism is proposed to achieve the efficient utilization of solvers. Extensive experiments have been conducted on synthetic instances of varying scales to investigate the efficiency of ALAHH. The experimental results demonstrate that the objective values and the computation time are significantly lower than those of SA and LAHC, and our algorithm has achieved the best solutions for 16 problems to date. Additionally, the impacts of two key parameters and mechanisms have been analyzed, and further validation of the robustness of the algorithm parameters and the effectiveness of the mechanisms.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129810"},"PeriodicalIF":7.5,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221875","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":"Task-free continual generative modelling via dynamic teacher-student framework","authors":"Fei Ye ,&nbsp;Adrian G. Bors","doi":"10.1016/j.eswa.2025.129873","DOIUrl":"10.1016/j.eswa.2025.129873","url":null,"abstract":"<div><div>Continually learning and acquiring new concepts from a dynamically changing environment is an important requirement for an artificial intelligence system. However, most existing deep learning methods fail to achieve this goal and suffer from significant performance degeneration under continual learning. We propose a new unsupervised continual learning framework combining Long- and Short-Term Memory management for training deep learning generative models. The former memory system employs a dynamic expansion model (Teacher), while the latter uses a fixed-capacity memory buffer to store the update-to-date information. A novel Teacher model expansion approach, called the Knowledge Incremental Assimilation Mechanism (KIAM) is proposed. KIAM evaluates the probabilistic distance between the already accumulated information and that from the Short Term Memory (STM). The proposed KIAM adaptively expands the Teacher’s capacity and promotes knowledge diversity among the Teacher’s experts. As Teacher experts, we consider generative deep learning models such as : the Variational Autocencoder (VAE), the Generative Adversarial Network (GAN) or the Denoising Diffusion Probabilistic Model (DDPM). We also extend the KIAM-based model to a Teacher-Student framework in which we use a data-free Knowledge Distillation (KD) process to train a VAE-based Student without using any task information. The results on Task Free Continual Learning (TFCL) benchmarks show that the proposed approach outperforms other models.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129873"},"PeriodicalIF":7.5,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145268544","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":"STKG-TP: Depression recognition via spatial-temporal knowledge graph and trajectory-semantic cross-fusion with EEG signals","authors":"Chen Huang ,&nbsp;Huijie Liu ,&nbsp;Jiannong Cao ,&nbsp;Yan Zhang ,&nbsp;Chao Yang ,&nbsp;Jianhua Song ,&nbsp;Zhifei Li ,&nbsp;Xiaoyong Yan","doi":"10.1016/j.eswa.2025.129744","DOIUrl":"10.1016/j.eswa.2025.129744","url":null,"abstract":"<div><div>EEG signals carry important neurocognitive information for depression recognition. However, existing EEG-based depression recognition research faces challenges in addressing semantic interpretability and improving model robustness. Consequently, in this paper, to overcome these challenges, we propose STKG-TP, a novel depression recognition model that integrates Spatiotemporal Knowledge Graphs (STKG) with trajectory-semantic cross-fusion. Specifically, we design an STKG module to learn brain region activation patterns associated with different depressive states and construct a spatiotemporal knowledge graph to enhance the model’s generalization and robustness. In addition, we introduce a Trajectory Prompting module that transforms EEG signal trajectories into a structured semantic library, enabling neurocognitive interpretability at the semantic level. Extensive experimental evaluations on three publicly available EEG datasets demonstrate the superior performance of STKG-TP in addressing these challenges. Compared with existing state-of-the-art depression recognition models, STKG-TP improves Accuracy by 1.13 %, 0.61 %, and 1.29 %, and Kappa score by 3.14 %, 1.88 %, and 2.75 %, respectively. The STKG-TP code is publicly available at: <span><span>https://github.com/xuxuanya-love/STKG-TP</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 129744"},"PeriodicalIF":7.5,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145271449","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":"Truck scheduling optimization at a cold chain cross-docking terminal considering uncertainties and the door-mixed service mode","authors":"Feifeng Zheng ,&nbsp;Yuzhi Yi ,&nbsp;Ming Liu ,&nbsp;Huaxin Qiu","doi":"10.1016/j.eswa.2025.129849","DOIUrl":"10.1016/j.eswa.2025.129849","url":null,"abstract":"<div><div>The increasing global demand for perishable agricultural products necessitates advancements in cold chain logistics. Cross-docking, known for its efficiency, is particularly well-suited for the transfer and distribution of such goods. However, truck scheduling at cold chain cross-dock terminals (CDTs) presents unique challenges, including product perishability, stringent time windows, and temperature-controlled environments. This work investigates a truck scheduling problem within a cold chain CDT, explicitly addressing uncertainties in refrigerated product damage (affecting supply) and repackaging times. A two-stage stochastic programming model is developed to capture these uncertainties. To solve this model, a scenario reduction approach employing K-means++ and K-medoids clustering is used, followed by Sample Average Approximation. Small-scale instances are solved optimally using CPLEX. For larger instances, a novel hybrid heuristic algorithm, combining the global search capabilities of Genetic Algorithms with the local search capabilities of Adaptive Large Neighborhood Search and Simulated Annealing, is proposed. Numerical experiments demonstrate the effectiveness of this algorithm, and sensitivity analysis provides valuable managerial insights.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129849"},"PeriodicalIF":7.5,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221872","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}