Expert Systems with Applications最新文献

{"title":"Evolutionary constrained optimization based on causal random forest","authors":"Yinghan Hong ,&nbsp;Sirui Liang ,&nbsp;Jiahao Lian ,&nbsp;Guizhen Mai ,&nbsp;Liang Zhao ,&nbsp;Yueting Xu ,&nbsp;Yi Xiang ,&nbsp;Hao Zhang ,&nbsp;Fangqing Liu ,&nbsp;Zhifeng Hao","doi":"10.1016/j.eswa.2026.131417","DOIUrl":"10.1016/j.eswa.2026.131417","url":null,"abstract":"<div><div>Constrained optimization problems constitute a class of optimization tasks that aim to maximize or minimize an objective function subject to intricate constraints. Evolutionary algorithms are extensively employed to tackle these problems, but the inherent nonlinearity, discontinuity, and restricted feasible regions of constrained optimization problems present significant challenges, and existing approaches often rely on predefined rules or empirical thresholds, which limit their adaptability and hinder causal interpretability. To overcome this limitation, this study proposes an evolutionary constrained optimization algorithm based on causal random forest that leverages causal random forest to quantify the causal strength between objective optimization and constraint satisfaction, thereby guiding the evolutionary search in a principled and informed manner. Furthermore, a dynamic adaptive strategy-switching mechanism is incorporated into the algorithm to reduce the reliance on empirical thresholds and fixed rules, which enhances the self-adaptive capability of the algorithm under complex and sophisticated constraints. Extensive experimental results on the CEC2006, CEC2010, and CEC2017 benchmark suites demonstrate that the proposed method consistently outperforms existing methods, underscoring its effectiveness and robustness in handling constrained optimization problems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131417"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192850","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":"Using mobile charging stations as probes to discover latent EV charging demand in stochastic environments: A deep reinforcement learning approach","authors":"Atefeh Hemmati Golsefidi ,&nbsp;Frederik Boe Hüttel ,&nbsp;Samitha Samaranayake ,&nbsp;Francisco Câmara Pereira","doi":"10.1016/j.eswa.2026.131433","DOIUrl":"10.1016/j.eswa.2026.131433","url":null,"abstract":"<div><div>The increasing adoption of electric vehicles underscores the urgent need for efficient and reliable charging infrastructure. Fixed charging stations are crucial for meeting surging demand and ensuring convenient access, yet accurately predicting demand remains highly challenging, because charging patterns change in response to where stations are placed, which creates a cyclical dilemma for planning. However, mobile charging stations (MCSs) offer a novel solution by flexibly relocating across urban areas, they can both deliver energy and act as dynamic probes to collect real-time data on charging demand. Existing studies, however, typically assume prior knowledge of demand distributions, which is rarely available in emerging EV markets or where privacy concerns limit data access. This paper proposes a Deep Reinforcement Learning (DRL) approach, formulated as a Partially Observable Markov Decision Process (POMDP), to optimize the relocation of MCSs in conjunction with fixed charging stations, while simultaneously uncovering latent demand patterns. We employ an Advantage Actor-Critic (A2C) algorithm with Long Short-Term Memory (LSTM) networks to capture temporal dependencies and adapt to stochastic demand. A dynamic Mixed-Integer Programming (MIP) model is developed as a benchmark that represents an idealized case with perfect foresight of demand. We compare the DRL agent against this optimization model in two settings: (i) a synthetic toy environment for controlled testing, and (ii) a realistic simulation of the Frederiksberg municipality in Denmark, calibrated with real charging data. The results show that the DRL framework effectively adapts to stochastic demand, outperforms the optimization baseline under uncertainty, and scales efficiently to larger problem instances. Beyond methodological contributions, the findings highlight how MCSs can serve a dual role as infrastructure supplements and as demand-discovery tools, offering valuable insights for data-driven and adaptive EV charging planning.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131433"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191899","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":"AnomalyLVM:Vision-language models for zero-shot anomaly detection","authors":"Yuqing Zhao,&nbsp;Min Meng,&nbsp;Jigang Wu","doi":"10.1016/j.eswa.2026.131392","DOIUrl":"10.1016/j.eswa.2026.131392","url":null,"abstract":"<div><div>Zero-Shot Anomaly Detection (ZSAD) has emerged as a promising approach for identifying unseen defects without requiring annotated training samples, but existing methods typically focus only on image-level detection and overlook fine-grained pixel-level localization. To address this gap, we propose AnomalyLVM, a unified vision-language framework designed to simultaneously handle image-level classification and pixel-level segmentation in zero-shot settings. AnomalyLVM leverages frozen SAM2 and DINO-X as dual visual encoders to extract complementary spatial and semantic features, which are fused and decoded via a lightweight decoder to generate localization maps. Meanwhile, a frozen CLIP text encoder guides image-level detection through semantic similarity matching. To enhance the accuracy of pixel-wise supervision, we introduce a Feature Enhancement Module (FEM) that dynamically refines static LayerCAM maps by integrating attention cues from both visual encoders and decoder affinity signals, resulting in more consistent and context-aware pseudo labels. Additionally, we adopt a prompt-free, object-agnostic strategy that replaces handcrafted templates with learnable, generic prompts, enabling AnomalyLVM to generalize across diverse categories and defect types without relying on domain-specific knowledge. Extensive experiments conducted across 17 real-world anomaly detection datasets from industrial and medical domains indicate that AnomalyLVM outperforms other ZSAD methods and can generalize better to different categories and even domains. Code will be made available at <span><span>https://github.com/hanli6688/AnomalyLVM</span><svg><path></path></svg></span></div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131392"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191893","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":"DSF-Unet: A dual-stream fusion denoising diffusion framework for imbalanced wafer defect classification","authors":"Rongbin Xu ,&nbsp;Zhiqiang Xu ,&nbsp;Jixiang Wang ,&nbsp;Ying Xie ,&nbsp;Lijie Wen ,&nbsp;Yun Yang","doi":"10.1016/j.eswa.2026.131471","DOIUrl":"10.1016/j.eswa.2026.131471","url":null,"abstract":"<div><div>Wafer defect recognition is crucial for semiconductor manufacturing. However, its accuracy is often limited by severe imbalance among defect categories. To address this challenge, we propose a Dual-Stream Fusion Unet denoising diffusion framework (DSF-Unet), which synthesizes diverse and high-quality wafer samples for minority classes. DSF-Unet builds upon a Unet encoder-decoder backbone and introduces two complementary components. The Bidirectional Mamba (Bi-Mamba) module models long-range spatial dependencies through state-space dynamics, while the Adaptive multi-scale attention (Am-att) module enhances spatial feature representation via cross-channel calibration. By jointly capturing global contextual information and fine-grained local defect patterns, these two modules enable the generation of balanced and representative synthetic samples, thereby effectively alleviating data imbalance. For further performance improvement, an enhanced Channel-Spatial Residual Network (CS-ResNet) is introduced for classification, which embeds a dual channel-spatial attention mechanism into the ResNet backbone to recalibrate feature responses and highlight defect-relevant regions. Experiments on two benchmark datasets demonstrate that DSF-Unet achieves superior performance, reaching 95.12% accuracy on WM-811K and 98.25% on Mixed-WM38.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131471"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192582","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":"Escaping from saddle points with perturbed gradient estimation","authors":"Jingjing Chen ,&nbsp;Sanyang Liu","doi":"10.1016/j.eswa.2026.131549","DOIUrl":"10.1016/j.eswa.2026.131549","url":null,"abstract":"<div><div>For non-convex functions where derivative information is difficult to obtain, escaping saddle points remains a significant challenge. Existing zeroth-order optimization algorithms approximate the true gradient using unbiased gradient estimation techniques, employing zero-mean random perturbations, or exploring negative curvature directions to escape saddle points. However, these methods encounter near-zero approximate gradients in the vicinity of saddle points, necessitating multiple small perturbations to escape, thereby consuming a substantial number of function evaluations. In this work, we propose the Two-step Simultaneous Perturbation Stochastic Approximation (2-SPSA) approach, to facilitate saddle point escape, which requires fewer function evaluations. At each iteration, this method requires only 4 function evaluations to estimate the gradients at the current point and its neighboring point, of which their convex combination serves as the descent direction. The randomness inherent in this gradient estimation aids in rapidly jumping out of saddle points. Experimental results indicate that the proposed method can escape saddle points with fewer function evaluations compared to other zeroth-order optimization algorithms.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131549"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192557","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":"Bayesian inference of nonlinear malaria dynamics in Ghana via an ensemble Markov chain Monte Carlo sampler","authors":"T. Ansah-Narh ,&nbsp;Y. Asare Afrane ,&nbsp;J. Bremang Tandoh","doi":"10.1016/j.eswa.2026.131540","DOIUrl":"10.1016/j.eswa.2026.131540","url":null,"abstract":"<div><div>Reliable quantification of malaria dynamics in sub-Saharan Africa remains hindered by short, noisy, and spatially heterogeneous surveillance records that challenge the assumptions of conventional deterministic models. In Ghana, health-facility data between 2014 and 2023 reveal highly non-linear and age-specific fluctuations in hospital admissions, yet existing approaches struggle to capture such stochastic variability or provide credible uncertainty bounds. This study develops a Bayesian nonlinear inference framework that integrates a cubic deterministic baseline with a damped oscillatory kernel, estimated via an affine-invariant ensemble Markov Chain Monte Carlo sampler. The framework accommodates limited data, explicitly models parameter uncertainty, and generates probabilistic forecasts of malaria admissions for children under five years and individuals aged five years or more. Results demonstrate that the proposed hybrid cubic-damped oscillatory kernel model achieves strong empirical adequacy (<span><math><mrow><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>0.9958</mn></mrow></math></span> for  &lt; 5 years; <span><math><mrow><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>0.9956</mn></mrow></math></span> for  ≥ 5 years) with residual errors below 2% and unimodal, well-mixed posterior distributions confirming robust convergence. District-level analysis reveals pronounced spatial heterogeneity, with coefficients of variation ranging from  &lt; 0.07 in stable urban centres such as Kumasi to  &gt; 3.3 in peripheral districts including Mpohor and Bia East. Forecasts for 2024–2026 indicate a gradual resurgence in admissions, increasing from approximately 137,000 to 149,000 cases among children under five and from 348,000 to 375,000 among older individuals, with uncertainty widening modestly over time. By producing interpretable probabilistic forecasts, this Bayesian framework provides a principled tool for anticipating short-term malaria fluctuations, guiding resource allocation, and strengthening data-driven decision-making within Ghana’s national malaria control strategy.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131540"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192714","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 hybrid FMECA framework integrating fuzzy–Bayesian reasoning for modeling complex risk dependencies in safety-critical systems","authors":"Akbar Rostamabadi ,&nbsp;Esmaeil Zarei","doi":"10.1016/j.eswa.2026.131530","DOIUrl":"10.1016/j.eswa.2026.131530","url":null,"abstract":"<div><div>Failure Modes, Effects, and Criticality Analysis (FMECA) is a widely used tool for risk and reliability assessment of engineering systems. However, its applicability to complex safety–critical systems is limited by several shortcomings, including the neglect of unknown failure causes, the focus on single failure modes, and the reliance on crisp numbers. This study proposes an Integrated Fuzzy–BWM–BN–FMECA (IFBBF) framework to model complex risk dependencies and address uncertainties in the failure analysis of complex safety–critical systems. The model consists of three steps. First, a qualitative analysis framework is developed based on the FMECA structure. Second, the qualitative framework is mapped into a Bayesian Network (BN) to model dependencies within cause–failure–effect chains and enable probabilistic updating. Advanced soft-computing techniques—including the Fuzzy Best–Worst Method (F-BWM), Noisy-OR (N-OR), and Leaky Noisy-OR (LN-OR) logics—are applied to reduce uncertainty and account for unknown failure causes. Third, a criticality analysis is performed using fuzzy BN reasoning, applying both predictive and diagnostic inferences. The proposed model is applied to the failure analysis of an industrial fire-tube boiler. Results indicated that the proposed approach significantly enhances the criticality assessment of failure modes compared to the conventional Risk Priority Number (RPN). It provides a comprehensive analysis of cascading effects and interdependencies among failure-causes-effect chains, enables probabilistic updating, and incorporates unknown failure causes— issues that conventional FMECA and other fuzzy-MCDM-based FMECA models cannot adequately address. The study findings demonstrate substantial improvements in the analytical capability of FMECA, offering a more flexible, detailed, and reliable framework for risk analysis of complex safety–critical systems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131530"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192587","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":"Adaptive bottleneck transformer for multimodal EEG, audio, and vision fusion","authors":"Sabina Bralina ,&nbsp;Adnan Yazici ,&nbsp;Cuntai Guan ,&nbsp;Min-Ho Lee","doi":"10.1016/j.eswa.2026.131487","DOIUrl":"10.1016/j.eswa.2026.131487","url":null,"abstract":"<div><div>Facial and speech expressions are primary cues for emotion recognition, while EEG provides a complementary neural perspective when external signals are ambiguous or absent. Although each modality contributes unique affective information, integrating such heterogeneous signals remains a major challenge in multimodal fusion research. To address this, the Adaptive Multimodal Bottleneck Transformer (AMBT) is introduced as a novel architecture, enabling efficient cross-modal interaction through adapters embedded within intermediate Transformer layers. These adapters 1) enhance stability by leveraging bottleneck tokens to prevent premature collapse, 2) enrich backbone representations while preserving unimodal capacity, 3) enable seamless integration across heterogeneous Transformer architectures, and 4) enable parameter-efficient training with fewer than 1% additional trainable parameters. AMBT was evaluated on three benchmark datasets: EAV (85.1%), CREMA-D (90.9%), and DEAP (98.7%), demonstrating competitive performance across all datasets. This results demonstrate the ability of AMBT to exploit complementary multimodal signals in a computationally efficient manner.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131487"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192713","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 digital twin-based approach for dynamic traffic-aware routing and charging of electric vehicles","authors":"Yue Xie ,&nbsp;Shanshan Li ,&nbsp;Linjun Lu ,&nbsp;Yuandong Pan ,&nbsp;Fumiya Iida","doi":"10.1016/j.eswa.2026.131380","DOIUrl":"10.1016/j.eswa.2026.131380","url":null,"abstract":"<div><div>The growing adoption of electric vehicles (EVs) presents new challenges for intelligent transportation systems (ITS), particularly in dynamic traffic environments where routing and charging decisions must adapt to fluctuating conditions. This paper proposes a Digital Twin-based Electric Vehicle Routing and Charging approach (DT-EVRC) that integrates real-time traffic data, predictive analytics, and a Dual-Population Evolutionary Algorithm (DPEA) to optimize EV travel and charging schedules. Unlike traditional static or simplified models, DT-EVRC continuously synchronizes with the physical transportation network, capturing variations in traffic density, charging station availability, and energy constraints. Experimental results on diverse grid-based urban scenarios demonstrate that DT-EVRC achieves robust and adaptive performance under traffic disruptions, road closures, and charging station failures. The proposed approach highlights the potential of digital twin technologies, combined with advanced optimization, to support next-generation ITS by enabling efficient, resilient, and sustainable urban mobility.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131380"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192698","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":"Doctors ranking based on PSO-BP model and attention mechanism with variable weights from the perspective of online medical consultation platforms","authors":"Na Zhao ,&nbsp;Zihao Zhang ,&nbsp;Fengjun Liu ,&nbsp;Zeshui Xu ,&nbsp;Yongqing Yang","doi":"10.1016/j.eswa.2026.131349","DOIUrl":"10.1016/j.eswa.2026.131349","url":null,"abstract":"<div><div>Current online medical consultation platforms face the challenge of low doctor engagement, significantly hindering the service quality and sustainable development. Existing doctor ranking mechanisms fail to effectively motivate doctors’ active participation and fully harness their contribution potential. Therefore, this paper proposes a novel doctor ranking method from the perspective of online medical consultation platforms to stimulate doctors’ engagement. Specifically, this paper first constructs a comprehensive doctor evaluation attribute system based on incomplete online data. Then, we apply the particle swarm optimization-backpropagation model and an attention mechanism to calculate the initial weights of evaluation attributes for reflecting the influence of different attributes on doctor rankings. After that, based on doctors’ behaviors and performances, and their impact on platform ecology and patient experience, we obtain the variable weights by modifying the initial weights of contribution attributes through an incentive and penalty mechanism and a state variable weight function, thereby reflecting the impact of doctors’ contributions on attribute weights. Additionally, we employ a score function with variable weights to calculate doctors’ scores and rankings, highlighting the effect of doctors’ contributions on their rankings. The proposed doctor ranking method is validated with a real dataset of 131,721 messages about 11,539 cardiovascular doctors from the well-known Chinese online medical consultation platform, Haodf.com. Ranking results of ten randomly selected doctors show that six doctors experience ranking changes due to their higher or lower levels of contribution, confirming the method’s effectiveness. Sensitivity analysis and evaluations of accuracy and usability further validate the robustness and reliability. Comparative experiments with state-of-the-art large language models and baseline methods reveal that the proposed method captures the complex interactions and latent relationships within large-scale online medical data, and more directly and clearly reflects the influence of doctors’ different levels of contribution on their rankings, providing a distinct advantage in incentivizing or penalizing doctors with exceptional or subpar performance.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"312 ","pages":"Article 131349"},"PeriodicalIF":7.5,"publicationDate":"2026-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146192787","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}