Expert Systems最新文献

{"title":"Intelligent Task Offloading in MEC-Enabled IoV Systems: From Data-Driven Learning to Knowledge-Based Reasoning","authors":"Pham Anh Thu,&nbsp;Gwanggil Jeon,&nbsp;Nguyen Tuan Linh,&nbsp;Chu Thi Minh Hue,&nbsp;Doan Thi Xuan,&nbsp;Dang Van Anh","doi":"10.1111/exsy.70426","DOIUrl":"https://doi.org/10.1111/exsy.70426","url":null,"abstract":"<div>\u0000 \u0000 <p>The rapid growth of the Internet of Vehicles (IoV) has created an urgent need for low-latency, energy-efficient and intelligent task processing in highly dynamic vehicular environments. This paper presents a comprehensive review of intelligent offloading approaches for MEC-enabled IoV networks from an expert systems perspective. First, we systematically classify existing methods into four categories: heuristic-based, optimization-based, learning-based and hybrid intelligent approaches. A comparative analysis is then conducted based on key performance metrics, including latency, energy efficiency, scalability, adaptability and sustainability. Beyond conventional performance evaluation, this review highlights the growing importance of knowledge-driven decision-making, explainability and hybrid reasoning mechanisms in enhancing the transparency, reliability and trustworthiness of offloading decisions in safety-critical vehicular applications. Furthermore, the paper identifies key research challenges, including mobility awareness, real-time adaptability, privacy preservation, interoperability and carbon-aware edge intelligence. A conceptual framework integrating deep reinforcement learning with expert systems is also discussed to illustrate future directions toward explainable and sustainable offloading. Finally, open research opportunities are outlined, emphasizing the development of intelligent, adaptive and environmentally sustainable offloading frameworks for next-generation 6G-enabled IoV systems.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 11","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149047914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Few Shot Learning for Flame State Monitoring From Limited Visible and Infrared Images","authors":"Ruiyuan Kang,&nbsp;Dimitrios C. Kyritsis,&nbsp;Panos Liatsis","doi":"10.1111/exsy.70408","DOIUrl":"https://doi.org/10.1111/exsy.70408","url":null,"abstract":"<p>The success of machine learning in image-based combustion monitoring requires big data, which is costly, and even impossible in industrial applications. To address this, we introduce metric-based few shot learning for combustion monitoring and classification, to the best of our knowledge, for the first time. Two algorithms, that is, Siamese Network coupled with k Nearest Neighbors (SN-kNN) and Prototypical Network (PN), are proposed. We explore the use of both visible and infrared (IR) images. We analyse the training process, test performance and inference speed of two algorithms using these image modalities, while t-SNE is used to visualize the learned features. The results demonstrate that both SN-kNN and PN are capable of distinguishing flame states, by just using 20 images per state. It is found that using visible images offers the ability to distinguish more substantial differences between classes, and identify more consistent patterns within each class, leading to higher training speed and model performance. In contrast, the relatively low quality of IR images complicates the extraction of distinguishable prototypes, as is in the case of PN, however, nevertheless offering high classification rates (F1-score <span></span><math>\u0000 <semantics>\u0000 <mrow>\u0000 <mo>≈</mo>\u0000 </mrow>\u0000 <annotation>$$ approx $$</annotation>\u0000 </semantics></math>0.99 averaged over five random seeds). When using the entire training set, SN-kNN performed well with IR images. On the other hand, due to its architectural design, PN offers faster training speed and inference. To the best of our knowledge, this is the first application of few shot learning methods on both image modalities for combustion state monitoring. We further position few shot learning among alternative data-efficient learning paradigms, analyse the practical deployment trade-offs of the two models, provide a controlled ablation of the principal design choices, and quantify temporal generalization, showing that a temporally representative 20-image calibration set restores near-ceiling accuracy under a strict temporal split, thus offering guidance for their future utilization in combustion monitoring tasks.</p>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 11","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/exsy.70408","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149047849","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"TC-DNet: Decomposition-Centric Dual-Branch Network for Long-Term Quality Forecasting in Batch Processes","authors":"Xiaoqiang Zhao,&nbsp;Guangbo Yu,&nbsp;Yongyong Hui","doi":"10.1111/exsy.70430","DOIUrl":"https://doi.org/10.1111/exsy.70430","url":null,"abstract":"<div>\u0000 \u0000 <p>Industrial batch processes are characterised by strong multivariate coupling, complex nonlinear dynamics and long-range temporal dependencies, posing significant challenges to accurate quality prediction. To address these issues, this paper proposes a Trend–Cycle Decomposition Network (TC-DNet), with a decomposition-oriented dual-branch deep learning framework. First, a Dynamic Sparse Attention (DSA) mechanism is developed to improve the computational efficiency of long-sequence modelling through the dynamic selection of informative temporal regions. Second, a Multiscale Dynamic Convolutional Gating Module (MDCGM) is proposed to effectively capture multiscale local patterns and enhance feature representation. Third, a Hierarchical Adaptive Decomposition with Cross-Phase Attention (HAD-CPA) module is designed to explicitly model the interactions among trend, cyclical and residual components, thereby improving the stability of long-term forecasting. Finally, experimental results on the penicillin fermentation and semiconductor etching datasets demonstrate that TC-DNet achieves superior forecasting performance across multiple prediction horizons. Compared with state-of-the-art decomposition-based and Transformer-based methods, TC-DNet exhibits significant advantages in long-term forecasting tasks, achieving up to a 23.0% reduction in RMSE at the longest forecasting horizon (720 steps) and a 14.8% reduction at the medium forecasting horizon (336 steps), while maintaining robust performance across all evaluated prediction horizons.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 11","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149007520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Age-Aware Energy and Resource Optimization for Federated Air Quality Monitoring With UAVs","authors":"Ali Kadhum M. Al-Qurabat","doi":"10.1111/exsy.70411","DOIUrl":"https://doi.org/10.1111/exsy.70411","url":null,"abstract":"<div>\u0000 \u0000 <p>Reliable air quality monitoring in Internet of Things (IoT) systems requires a synergy between energy-efficient data collection, data freshness and scalable learning across geographically distributed sensors. We propose AERO-Air (Age-aware Energy and Resource Optimization), a novel hierarchical framework that jointly optimizes Unmanned Aerial Vehicle (UAV) trajectories, Age-of-Information (AoI)-aware client scheduling and hierarchical federated aggregation. By focusing on the temporal relevance of environmental data, AERO-Air improves prediction accuracy, information freshness and UAV energy efficiency. The framework leverages lightweight local neural models at ground sensors, performs intermediate aggregations at follower UAVs, and utilizes a leader UAV to coordinate global updates via hierarchical Federated Learning (FL). The optimization problem, which jointly considers UAV mobility, AoI constraints and federated learning objectives, is addressed through a hybrid solution combining offline successive convex approximation (SCA) and online reinforcement learning (RL). Simulation results on the University of California Irvine (UCI) Air Quality dataset confirm that AERO-Air significantly reduces AoI, accelerates learning convergence, and maintains high energy efficiency compared to existing baselines.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148942142","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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 Hybrid Feature Selection Method Based on Dual Attention Mechanism for High-Dimensional Data Classification","authors":"Ruixiang Wang,&nbsp;Yu Xue,&nbsp;Ilyas Muhammad,&nbsp;Moncef Gabbouj","doi":"10.1111/exsy.70398","DOIUrl":"https://doi.org/10.1111/exsy.70398","url":null,"abstract":"<div>\u0000 \u0000 <p>As an important preprocessing technique in the field of machine learning, feature selection serves the purpose of identifying the optimal feature subset, while eliminating irrelevant or redundant features. However, most existing attention mechanism-based feature selection methods only consider promising features when handling high-dimensional datasets, which may lead to redundancy of promising features and neglect of the interaction between irrelevant features and promising features. So it remains a challenge to effectively screen out features that are complementary to the promising ones from irrelevant features. To solve this limitation, this paper proposes a hybrid feature selection method based on dual attention mechanism. First, an initialization method based on the knee-point strategy is proposed to initialize the dataset into a main dataset and an auxiliary dataset. Then, a training module and a two-stage optimization are used to generate the main feature set and auxiliary feature set which include the elite features. Finally, a complementary mining operation is employed to screen out the optimal feature subset by leveraging the correlation between the two aforementioned feature sets. Experimental evaluations conducted on 24 benchmark datasets verify that the proposed algorithm outperforms eight state-of-the-art algorithms in terms of both classification accuracy and selected feature size, thereby illustrating its effectiveness in solving practical problems.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148943745","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Cross-Lingual Knowledge Transfer Framework for Multilingual Customer Support AI: An Expert System Perspective","authors":"Joona Mäntyvaara,&nbsp;Paavo Nevalainen,&nbsp;Jukka Heikkonen","doi":"10.1111/exsy.70414","DOIUrl":"https://doi.org/10.1111/exsy.70414","url":null,"abstract":"<p>This study presents a comprehensive methodology for processing multilingual customer support data to prepare it for training AI-based conversational systems. Using a dataset of 36,599 unique customer interactions from a Finnish energy company, we employed language-specific BERT models to identify and analyse thematic patterns within customer inquiries. Our dual-model approach separately processed Finnish and English content, utilizing specialized embeddings for each language before comparing topics across languages using cosine similarity measures. Our preprocessing pipeline employs confidence-weighted language detection optimized for technical domain terminology, adaptive frequency filtering with domain preservation and morphology-aware language processing to address the unique challenges of multilingual customer support data. Results revealed distinct topic distributions between languages with strategically important overlapping themes, particularly in technical domains where similarity coefficients reached 0.700. UMAP visualizations confirmed the computational findings, showing well-defined clusters in Finnish content and more diffuse boundaries in English topics. The identified cross-language similarities in contract and energy-related topics provide promising opportunities for targeted knowledge transfer, while disparities in other domains highlight the need for language-specific optimization. The methodology presented demonstrates a robust framework for analysing bilingual customer support corpora that can be extended to training specialized AI assistants while addressing language-specific nuances.</p>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/exsy.70414","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148943501","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SCAgent: An LLM-Based Agent With State-Aware Coordination for Automated Program Repair","authors":"Shaosheng Wang,&nbsp;Lu Lu,&nbsp;Shaojian Qiu","doi":"10.1111/exsy.70400","DOIUrl":"https://doi.org/10.1111/exsy.70400","url":null,"abstract":"<div>\u0000 \u0000 <p>Automated program repair (APR) has emerged as a promising technique for improving software reliability by automatically fixing bugs. Recent advances in large language models (LLMs), particularly LLM-based agents, have further enhanced repair capabilities by enabling flexible patch generation and iterative reasoning. However, existing LLM-based agent frameworks still suffer from several limitations, including unstructured repair processes, unproductive trial-and-error loops and the lack of explicit coordination between reasoning and action, which lead to inefficient and unstable repair behaviours. In this paper, we propose SCAgent, an LLM-based agent for APR with state-aware coordination. SCAgent structures the repair process into coordinated stages, including bug understanding, information gathering and patch generation, and dynamically controls transitions between these stages based on the repair context. Experimental results on Defects4J show that SCAgent correctly repairs 173 bugs. We further conduct a preliminary evaluation on GitBug-Java to assess its potential generalizability, with SCAgent correctly repairing 15 bugs in a sampled subset of 100 bugs.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148849318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"Ensembling Heterogeneous NLU Models via Knowledge-Driven Constraints for Task-Oriented Dialogue Systems","authors":"Yuyan Wu,&nbsp;Romina Soledad Albornoz-De Luise,&nbsp;Miguel Arevalillo-Herráez","doi":"10.1111/exsy.70404","DOIUrl":"https://doi.org/10.1111/exsy.70404","url":null,"abstract":"<div>\u0000 \u0000 <p>Intent classification and slot filling are essential tasks in natural language understanding. The latest approaches consider these tasks jointly in order to leverage their interdependence but use single models for training and inference, which struggle to capture fine semantic distinctions and contextual dependencies in diverse input patterns. Recently, the integration of expert-defined constraints into these models has gained traction as a promising approach, effectively bridging the gap between data-driven methodologies and expert knowledge. In this paper, we advance this hybrid intelligence approach by employing model ensembles that combine the outputs from two models using expert-defined constraints. The proposed approach was evaluated using combinations of four state-of-the-art models (Bi-model, DCA-Net, E2EMG-CRF and AGIF) across three widely used benchmark datasets (ATIS, SNIPS and NLU-Benchmark). The results consistently indicate substantial improvements in all metrics, including intent accuracy, slot F1-score and semantic accuracy. The magnitude of the improvement depends on the dataset and ensemble configuration, with average absolute improvements in semantic accuracy ranging from approximately 0.7% to 1.8% points across the evaluated datasets and the largest improvement exceeding 6% points.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148848901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"When Content Signals Lose Credibility: Signalling Equilibria on Human-Centred Platforms in the LLM Era","authors":"Miao He,&nbsp;Weiwei Jiang,&nbsp;Momina Shaheen,&nbsp;Haihua Xie,&nbsp;Mingming Sun,&nbsp;Rajkumar Singh Rathore,&nbsp;Weixi Gu","doi":"10.1111/exsy.70403","DOIUrl":"https://doi.org/10.1111/exsy.70403","url":null,"abstract":"<div>\u0000 \u0000 <p>On platforms where readers value personal voice and human-authored expression, fluent writing used to be a credible sign that a post reflected real human effort. Large language models have made such writing cheap to produce, so readers can no longer reliably infer whether a post is human-authored from surface quality alone. We develop a signalling framework to study how the spread of large language models (LLMs) can reshape content valuation on online platforms that focus on personal voice, interpersonal resonance, and the credibility of human-authored expression. Using Reddit data from a narrative, emotion-oriented community, we recover reduced-form sign patterns that discipline the post-LLM regime analysed in the model. We show that, in the model, asymmetric technological change can erode the credibility of conventional writing cues and move equilibrium from separating to semi-separating and, in some parameter regions, to pooling or collapse on this class of platforms. We then use agent-based simulations to compare four platform interventions. The simulations indicate that selectively increasing the cost borne by content producers whose output is more heavily dependent on AI generation provides the strongest benchmark performance in the model, while labelling with appeal offers a more attractive practical balance among the more implementable interventions.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148849176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"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":"SCRSNet: An Efficient Crowd Counting via Lightweight Spatial-Channel Reconstructive and Scale-Aware Network","authors":"Lin Qi,&nbsp;Jing-an Cheng,&nbsp;Yanfei Dong,&nbsp;Chenguang Wu,&nbsp;Zhao Huang,&nbsp;Mingliang Gao","doi":"10.1111/exsy.70409","DOIUrl":"https://doi.org/10.1111/exsy.70409","url":null,"abstract":"<div>\u0000 \u0000 <p>Accurate crowd counting is of paramount importance for urban management and public security. Currently, to pursue superior counting precision, mainstream methods tend to construct cumbersome models with a massive number of parameters. Although such progress undeniably boosts counting performance, the substantial parameter and heavy computational overhead of these models severely impede their deployment on resource-limited hardware, such as drones and mobile edge computing nodes. Lightweight architectures have surfaced as tailored solutions for edge computing, aiming to strike a compromise between inference speed and predictive precision. Nevertheless, these lightweight models often falter when confronted with real-world visual complexities, notably intricate background interference (e.g., architectural occlusions, cross-modal noise) and severe scale variations. To overcome these limitations, we propose the Spatial-Channel Reconstructive and Scale-aware Network for Crowd Counting (SCRSNet). Adopting a divide-and-conquer strategy, the proposed framework disentangles the latent representation into spatial and channel dimensions, attentively modelling their inter-dependencies to filter out crowd-irrelevant features. Specifically, a Spatial and Channel Reconstructive (SCR) module is constructed to mitigate the adverse effects of background noise. Concurrently, a Dilated Scale-aware Attention (DSA) module is introduced to capture multi-scale features by aggregating semantic details across varying scales. Extensive experiments on five crowd counting benchmarks demonstrate that the proposed SCRSNet surpasses existing state-of-the-art methods in both accuracy and efficiency.</p>\u0000 </div>","PeriodicalId":51053,"journal":{"name":"Expert Systems","volume":"43 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2026-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148849221","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}