Search Results (2,720)

Chronic obstructive pulmonary disease (COPD) is a major health burden worldwide and in Taiwan, ranking as the third leading cause of death globally, and its prevalence in Taiwan continues to rise. Readmission within 14 days is a key indicator of disease instability and care efficiency, driven jointly by patient-level physiological vulnerability (such as reduced lung function and multiple comorbidities) and healthcare system-level deficiencies in transitional care. To mitigate the growing burden and improve quality of care, it is urgently necessary to develop an AI-based prediction model for 14-day readmission. Such a model could enable early identification of high-risk patients and trigger multidisciplinary interventions, such as pulmonary rehabilitation and remote monitoring, to effectively reduce avoidable early readmissions. However, medical data are commonly characterized by severe class imbalance, which limits the ability of conventional machine learning methods to identify minority-class cases. In this study, we used real-world clinical data from multiple hospitals in Kaohsiung City to construct a prediction framework that integrates data generation and ensemble learning to forecast readmission risk among patients with chronic obstructive pulmonary disease (COPD). CTGAN and kernel density estimation (KDE) were employed to augment the minority class, and the impact of these two generation approaches on model performance was compared across different augmentation ratios. We adopted a stacking architecture composed of six base models as the core framework and conducted systematic comparisons against the baseline models XGBoost, AdaBoost, Random Forest, and LightGBM across multiple recall thresholds, different feature configurations, and alternative data generation strategies. Overall, the results show that, under high-recall targets, KDE combined with stacking achieves the most stable and superior overall performance relative to the baseline models. We further performed ablation experiments by sequentially removing each base model to evaluate and analyze its contribution. The results indicate that removing KNN yields the greatest negative impact on the stacking classifier, particularly under high-recall settings where the declines in precision and F1-score are most pronounced, suggesting that KNN is most sensitive to the distributional changes introduced by KDE-generated data. This configuration simultaneously improves precision, F1-score, and specificity, and is therefore adopted as the final recommended model setting in this study. Full article

Crop production in the desert Southwest of the United States, as well as in other arid and semi-arid regions, requires tools that provide accurate crop evapotranspiration (ET) estimates to support efficient irrigation management. Such tools include the web-based OpenET platform, which provides real-time ET data generated from six satellite-based models, their Ensemble, and a field-based system (LI-710, LI-COR Inc., Lincoln, NE, USA). This study evaluated simulated ET (ET_SIM) of cotton (Gossypium hirsutum L.) derived from OpenET models (ALEXI/DisALEXI, eeMETRIC, geeSEBAL, PT-JPL, SIMS, and SSEBop), their Ensemble approach, and LI-710. Field data were utilized to estimate cotton ET using the soil water balance (SWB) method (ET_SWB) from June to October 2025 in Gila Bend, AZ, USA. Four evaluation metrics, the normalized root-mean-squared error (NRMSE), mean bias error (MBE), simulation error (S_e), and coefficient of determination (R²), were employed to evaluate the performance of OpenET models, their Ensemble, and the LI-710 in estimating cotton ET. Statistical analysis indicated that the ALEXI/DisALEXI, geeSEBAL, and PT-JPL models substantially underestimated ET_SWB, with simulation errors ranging from −26.92% to −20.57%. The eeMETRIC, SIMS, SSEBop, and Ensemble provided acceptable ET estimates (22.57% ≤ NRMSE ≤ 29.85%, −0.36 mm. day⁻¹ ≤ MBE ≤ 0.16 mm. day⁻¹, −7.58% ≤ S_e ≤ 3.42%, 0.57 ≤ R² ≤ 0.74). Meanwhile, LI-710 simulated cotton ET acceptably with a slight tendency to overestimate daily ET by 0.21 mm. A strong positive correlation was observed between daily ET_SIM from LI-710 and ET_SWB, with S_e and NRMSE of 4.40% and 23.68%, respectively. Based on our findings, using a singular OpenET model, such as eeMETRIC, SIMS, or SSEBop, the OpenET Ensemble, and the LI-710 can offer growers and decision-makers reliable guidance for efficient irrigation management of late-planted cotton in arid and semi-arid climates. Full article

Endometrial lesions are common in gynecology, exhibiting considerable clinical heterogeneity across different subtypes. Although ultrasound imaging is the preferred diagnostic modality due to its noninvasive, accessible, and cost-effective nature, its diagnostic performance remains highly operator-dependent, leading to subjectivity and inconsistent results. To address these limitations, this study proposes a hierarchical feature selection and optimization framework for endometrial lesions, aiming to enhance the objectivity and robustness of ultrasound-based diagnosis. Firstly, Kernel Principal Component Analysis (KPCA) is employed for nonlinear dimensionality reduction, retaining the top 1000 principal components. Secondly, an ensemble of three filter-based methods—information gain, chi-square test, and symmetrical uncertainty—is integrated to rank and fuse features, followed by thresholding with Maximum Scatter Difference Linear Discriminant Analysis (MSDLDA) for preliminary feature selection. Finally, the Whale Migration Algorithm (WMA) is applied to population-based feature optimization and classifier training under the constraints of a Support Vector Machine (SVM) and a macro-averaged F1 score. Experimental results demonstrate that the proposed closed-loop pipeline of “kernel reduction—filter fusion—threshold pruning—intelligent optimization—robust classification” effectively balances nonlinear structure preservation, feature redundancy control, and model generalization, providing an interpretable, reproducible, and efficient solution for intelligent diagnosis in small- to medium-scale medical imaging datasets. Full article

The stability of ice structures in cold regions and polar environments has been increasingly challenged by global warming and climate change, making the accurate estimation of ice flexural properties essential. However, the flexural failure process of ice is highly complex, and the calculated flexural properties are influenced by multiple factors. Hence, several data-driven artificial intelligence models were developed to predict flexural strength, using classification and regression tree (CART), AdaBoost, and Random Forest methods, while the Elitist Ant System (EAS) was applied to optimize model parameters. The EAS procedure converged rapidly within ten iterations and effectively enhanced overall model performance. Compared with the single CART model, ensemble approaches exhibited higher prediction accuracy and better generalization, with AdaBoost achieving the best performance (R² = 0.736). Feature-importance analysis indicated that the testing method and specimen geometry had the greatest influence on the results, highlighting the importance of careful control of experimental conditions. The proposed ensemble–metaheuristic framework provides an efficient tool for predicting the mechanical behavior of ice and offers useful support for stability assessments of ice structures under changing climatic conditions. Full article

The rapid deterioration of global infrastructure necessitates precise and automated crack detection technologies for proactive maintenance. However, deep learning-based segmentation models often suffer from a scarcity of diverse, high-quality labeled datasets. This study proposes StyleSPADE, a novel conditional image generation model that integrates semantic masks and style images to synthesize realistic crack data with diverse background textures while preserving precise geometric morphology. To validate the effectiveness of the generated data, we conducted extensive semantic segmentation tasks using Transformer-based (Mask2Former, Swin-UPerNet) and CNN-based (K-Net) models. Experimental results demonstrate that StyleSPADE-based augmentation significantly outperforms baseline models, achieving a Crack IoU of 0.6376 and an F1-score of 0.7586. Furthermore, we implemented a Stacking Ensemble strategy combining high-recall and high-precision models, which further improved performance to a Crack IoU of 0.6452. Our findings confirm that StyleSPADE effectively mitigates the data scarcity problem and enhances the robustness of crack detection in complex environmental conditions. This framework contributes to improving the efficiency and safety of infrastructure management by enabling reliable damage assessment in data-limited environments. Full article

In recent years, deformation disasters in mine shafts have occurred frequently, posing a threat to mine safety. The nonlinear coupling relationship between shaft surrounding rock deformation and rock mass mechanical parameters is a key criterion for surrounding rock stability. However, existing machine learning prediction methods are rarely applied to shaft deformation, and issues such as poor accuracy and generalization of single models remain. To address this, the study proposes a feature-weighted Stacking ensemble model, which considers 15 feature variables; using RMSE, MAE, R², and inter-model MAPE correlation as evaluation metrics, GBDT, XGBoost, KNN, and MLP are selected as base learners, with Lasso linear regression as the meta-learner. Prediction errors are corrected by weighting the outputs of base learners based on prediction accuracy. Experiments show that, using MAPE as the evaluation metric, the improved model reduces the error by 2.59% compared with the best base learner KNN, by 6.83% compared with XGBoost, and by 0.18% more than the traditional Stacking algorithm, making it suitable for predicting weak surrounding rock shaft deformation under multi-feature conditions. Full article

This work presents the development and evaluation of Artificial Intelligence (AI) models for the automatic classification of brain tumors in Magnetic Resonance Imaging (MRI) scans. Several deep learning architectures were implemented and compared, including VGG-19, ResNet50, EfficientNetB3, Xception, MobileNetV2, DenseNet201, InceptionV3, Vision Transformer (ViT), and an Ensemble model. The models were developed in Python (version 3.12.4) using the Keras and TensorFlow frameworks and trained on a public Brain Tumor MRI dataset containing 7023 images. Data augmentation and hyperparameter optimization techniques were applied to improve model generalization. The results showed high classification performance, with accuracies ranging from 89.47% to 98.17%. The Vision Transformer achieved the best performance, reaching 98.17% accuracy, outperforming traditional Convolutional Neural Network (CNN) architectures. Explainable AI (XAI) methods Grad-CAM, LIME, and Occlusion Sensitivity were employed to assess model interpretability, showing that the models predominantly focused on tumor regions. The proposed approach demonstrated the effectiveness of AI-based systems in supporting early diagnosis of brain tumors, reducing analysis time and assisting healthcare professionals. Full article

We present a computational study that precedes the potential interactions between SARS-CoV-2 helicase (NSP13) and selected host proteins implicated in chaperone-assisted folding and polyamine metabolism. Using structure-based modelling and protein–protein docking (BioLuminate v4.6), followed by all-atom molecular dynamics (MD) simulations (GROMACS v2018.6), and comparative MM-GBSA scoring (HawkDock v2), we evaluated the stability and interface properties of NSP13 complexes with cytosolic heat shock proteins; heat shock protein 40 (HSP40), heat shock protein 70 (HSP70), heat shock protein 90 (HSP90) and the polyamine biosynthesis enzyme ornithine decarboxylase (ODC). Docking, MD, and interface analyses indicate distinct complex behaviours: HSP70-NSP13 complexes sampled compact conformations, HSP90-NSP13 ensembles displayed greater conformational heterogeneity but more favourable comparative MM-GBSA estimates, and ODC-NSP13 interfaces were comparatively well packed. Per-residue contact mapping identified a small set of recurrent NSP13 residues, Lys22 and Asn51, as putative interaction hotspots. The reported findings herein generate testable hypotheses about NSP13 recruitment of host chaperones and modulation of polyamine metabolism that may inform downstream experimental studies. Full article

With the acceleration of digital transformation, open-source software has become a fundamental component of modern smart grids and other critical infrastructures. However, the complex dependency structures of open-source ecosystems and the continuous emergence of vulnerabilities pose substantial challenges to software supply chain security. In power information networks and cyber–physical control systems, vulnerabilities in open-source components integrated into Supervisory Control and Data Acquisition (SCADA), Energy Management System (EMS), and Distribution Management System (DMS) platforms and distributed energy controllers may propagate along the supply chain, threatening system security and operational stability. In such application scenarios, large language models (LLMs) often suffer from limited semantic accuracy when handling domain-specific security terminology, as well as deployment inefficiencies that hinder their practical adoption in critical infrastructure environments. To address these issues, this paper proposes KD-SecBERT, a domain-specific semantic bidirectional encoder optimized through multi-level knowledge distillation for open-source software supply chain security in smart grid applications. The proposed framework constructs a hierarchical multi-teacher ensemble that integrates general language understanding, cybersecurity-domain knowledge, and code semantic analysis, together with a lightweight student architecture based on depthwise separable convolutions and multi-head self-attention. In addition, a dynamic, multi-dimensional distillation strategy is introduced to jointly perform layer-wise representation alignment, ensemble knowledge fusion, and task-oriented optimization under a progressive curriculum learning scheme. Extensive experiments conducted on a multi-source dataset comprising National Vulnerability Database (NVD) and Common Vulnerabilities and Exposures (CVE) entries, security-related GitHub code, and Open Web Application Security Project (OWASP) test cases show that KD-SecBERT achieves an accuracy of 91.3%, a recall of 90.6%, and an F1-score of 89.2% on vulnerability classification tasks, indicating strong robustness in recognizing both common and low-frequency security semantics. These results demonstrate that KD-SecBERT provides an effective and practical solution for semantic analysis and software supply chain risk assessment in smart grids and other critical-infrastructure environments. Full article

Data uncertainty and limited model generalization remain critical bottlenecks in large-scale remote sensing of soil salinization. Although the integration of multi-source data has improved predictive potential, conventional deterministic feature selection methods often overlook stochastic noise inherent in environmental variables, leading to models that overfit spurious correlations rather than learning stable physical signals. To address this limitation, this study proposes a Bootstrap–Boruta feature stability assessment framework that shifts feature selection from deterministic “feature importance” ranking to probabilistic “feature stability” evaluation, explicitly accounting for uncertainty induced by data perturbations. The proposed framework is evaluated by integrating stability-driven feature sets with multiple machine learning models, including a Back-Propagation Neural Network (BPNN) optimized using the Red-billed Blue Magpie Optimization (RBMO) algorithm as a representative optimization strategy. Using the Minqin Lake region as a case study, the results demonstrate that the stability-based framework effectively filters unstable noise features, reduces systematic estimation bias, and improves predictive robustness across different modeling approaches. Among the tested models, the RBMO-optimized BPNN achieved the highest accuracy. Under a rigorous bootstrap validation framework, the quality-controlled ensemble model yielded a robust mean R² of 0.657 ± 0.05 and an RMSE of 1.957 ± 0.289 dS/m. The framework further identifies eleven physically robust predictors, confirming the dominant diagnostic role of shortwave infrared (SWIR) indices in arid saline environments. Spatial mapping based on these stable features reveals that 30.7% of the study area is affected by varying degrees of soil salinization. Overall, this study provides a mechanism-driven, promising, within-region framework that enhances the reliability of remote-sensing-based soil salinity inversion under heterogeneous environmental conditions. Full article

Graph pseudo-labeling is an effective semi-supervised learning (SSL) approach to improve graph neural networks (GNNs) by leveraging unlabeled data. However, its success heavily depends on the reliability of pseudo-labels, which can often result in confirmation bias and training instability. To address these challenges, we propose a dual-layer consistency semi-supervised framework (DiPat), which integrates an internal differentiating pattern consistency mechanism and an external multimodal knowledge verification mechanism. In the internal layer, DiPat extracts multiple differentiating patterns from a single information source and enforces their consistency to improve the reliability of intrinsic decisions. During the supervised training phase, the model learns to extract and separate these patterns. In the semi-supervised learning phase, the model progressively selects highly consistent samples and ranks pseudo-labels based on the minimum margin principle, mitigating the overconfidence problem common in confidence-based or ensemble-based methods. In the external layer, DiPat also integrates large multimodal language models (MLLMs) as auxiliary information sources. These models provide latent textual knowledge to cross-validate internal decisions and introduce a responsibility scoring mechanism to filter out inconsistent or unreliable external judgments. Extensive experiments on multiple benchmark datasets show that DiPat demonstrates superior robustness and generalization in low-label settings, consistently outperforming strong baseline methods. Full article

Reliable operation of rolling bearings is essential for mechanical systems. Acoustic emission (AE) offers a promising approach for bearing fault detection because of its high-frequency response and strong noise-suppression capability. This study proposes an intelligent diagnostic method that combines an improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and a convolutional neural network–bidirectional long short-term memory (CNN–BiLSTM) architecture. The method first applies wavelet denoising to AE signals, then uses ICEEMDAN decomposition followed by kurtosis-based screening to extract key fault components and construct feature vectors. Subsequently, a CNN automatically learns deep time–frequency features, and a BiLSTM captures temporal dependencies among these features, enabling end-to-end fault identification. Experiments were conducted on a bearing acoustic emission dataset comprising 15 operating conditions, five fault types, and three rotational speeds; comparative model tests were also performed. Results indicate that ICEEMDAN effectively suppresses mode mixing (average mixing rate 6.08%), and the proposed model attained an average test-set recognition accuracy of 98.00%, significantly outperforming comparative models. Moreover, the model maintained 96.67% accuracy on an independent validation set, demonstrating strong generalization and practical application potential. Full article

Accurate short-term forecasting of photovoltaic (PV) output is critical to managing the variability of PV generation and ensuring reliable grid operation with high renewable integration. We propose an enhanced hybrid deep learning framework that combines Temporal Convolutional Networks (TCNs), Gated Recurrent Units (GRUs), and Random Forests (RFs) in an optimized weighted ensemble strategy. This approach leverages the complementary strengths of each component: TCNs capture long-range temporal dependencies via dilated causal convolutions; GRUs model sequential weather-driven dynamics; and RFs enhance robustness to outliers and nonlinear relationships. The model was evaluated on high-resolution operational data from the Yulara solar plant in Australia, forecasting horizons from 5 min to 1 h. Results show that the TCN-GRU-RF model consistently outperforms conventional benchmarks, achieving R² = 0.9807 (MAE = 0.0136; RMSE = 0.0300) at 5 min and R² = 0.9047 (RMSE = 0.0652) at 1 h horizons. Notably, the degradation in R² across forecasting horizons was limited to 7.7%, significantly lower than the typical 10–15% range observed in the literature, highlighting the model’s scalability and resilience. These validated results indicate that the proposed approach provides a robust, scalable forecasting solution that enhances grid reliability and supports the integration of distributed renewable energy sources. Full article

Surface deformation induced by the extraction of natural resources constitutes a non-stationary spatiotemporal process. Modeling surface deformation time series obtained through Interferometric Synthetic Aperture Radar (InSAR) technology using deep learning methods is crucial for disaster prevention and mitigation. However, the complexity of model hyperparameter configuration and the lack of interpretability in the resulting predictions constrain its engineering applications. To enhance the reliability of model outputs and their decision-making value for engineering applications, this study presents a workflow that combines a Tree-structured Parzen Estimator (TPE)-based Bayesian optimization approach with ensemble inference. Using the Rhineland coalfield in Germany as a case study, we systematically evaluated six deep learning architectures in conjunction with various spatiotemporal coding strategies. Pairwise comparisons were conducted using a Welch t-test to evaluate the performance differences across each architecture under two parameter-tuning approaches. The Benjamini–Hochberg method was applied to control the false discovery rate (FDR) at 0.05 for multiple comparisons. The results indicate that TPE-optimized models demonstrate significantly improved performance compared to their manually tuned counterparts, with the ResNet+Transformer architecture yielding the most favorable outcomes. A comprehensive analysis of the spatial residuals further revealed that TPE optimization not only enhances average accuracy, but also mitigates the model’s prediction bias in fault zones and mineralize areas by improving the spatial distribution structure of errors. Based on this optimal architecture, we combined the ten highest-performing models from the optimization stage to generate a quantile-based susceptibility map, using the ensemble median as the central predictor. Uncertainty was quantified from three complementary perspectives: ensemble spread, class ambiguity, and classification confidence. Our analysis revealed spatial collinearity between physical uncertainty and absolute residuals. This suggests that uncertainty is more closely related to the physical complexity of geological discontinuities and human-disturbed zones, rather than statistical noise. In the analysis of super-threshold probability, the threshold sensitivity exhibited by the mining area reflects the widespread yet moderate impact of mining activities. By contrast, the fault zone continues to exhibit distinct high-probability zones, even under extreme thresholds. It suggests that fault-controlled deformation is more physically intense and poses a greater risk of disaster than mining activities. Finally, we propose an engineering decision strategy that combines uncertainty and residual spatial patterns. This approach transforms statistical diagnostics into actionable, tiered control measures, thereby increasing the practical value of susceptibility mapping in the planning of natural resource extraction. Full article

Current transformer measurement errors exhibit strong non-stationarity and multi-scale temporal dynamics, which make accurate prediction challenging for conventional deep learning models. This paper presents a hybrid signal processing and temporal learning framework that integrates ensemble empirical mode decomposition (EEMD) with a dual-scale temporal convolutional architecture. EEMD adaptively decomposes the error sequence into intrinsic mode functions, while a Pearson correlation-based selection step removes redundant and noise-dominated components. The refined signal is then processed by a dual-scale temporal convolutional network (TCN) designed with parallel dilated kernels to capture both high-frequency transients and long-range drift patterns. Experimental evaluations on 110 kV substation data confirm that the proposed decomposition-enhanced dual-scale temporal convolutional framework significantly improves generalization and robustness, reducing the root mean square error by 40.9% and the mean absolute error by 37.0% compared with benchmark models. The results demonstrate that combining decomposition-based preprocessing with multi-scale temporal learning effectively enhances the accuracy and stability of non-stationary current transformer error forecasting. Full article