Search Results (6,390)

Fault detection and remaining useful life (RUL) prediction are critical tasks in self-healing network (SHN) environments and industrial cyber–physical systems. These domains demand intelligent systems capable of handling dynamic, high-dimensional sensor data. However, existing optimization-based approaches often struggle with imbalanced datasets, noisy signals, and delayed convergence, limiting their effectiveness in real-time applications. This study utilizes two benchmark datasets—EFCD and SFDD—which represent electrical and sensor fault scenarios, respectively. These datasets pose challenges due to class imbalance and complex temporal dependencies. To address this, we propose a novel hybrid framework combining Attention-Augmented Convolutional Neural Networks (AACNN) with transformer encoders, enhanced through Enhanced Ensemble-SMOTE for balancing the minority class. The model captures spatial features and long-range temporal patterns and learns effectively from imbalanced data streams. The novelty lies in the integration of attention mechanisms and adaptive oversampling in a unified fault-prediction architecture. Model evaluation is based on multiple performance metrics, including accuracy, F1-score, MCC, RMSE, and score*. The results show that the proposed model outperforms state-of-the-art approaches, achieving up to 97.14% accuracy and a score* of 0.419, with faster convergence and improved generalization across both datasets. Full article

The need to improve medical diagnosis is of utmost importance in medical research, consisting of the optimization of accurate classification models able to assist clinical decisions. To minimize the errors that can be caused by using a single classifier, the voting ensemble technique can be used, combining the classification results of different classifiers to improve the final classification performance. This paper aims to compare the existing voting ensemble techniques with a new game-theory-derived approach based on Shapley values. We extended this method, originally developed for binary tasks, to the multi-class setting in order to capture complementary information provided by different classifiers. In heterogeneous clinical scenarios such as thyroid nodule diagnosis, where distinct models may be better suited to identify specific subtypes (e.g., benign, malignant, or inflammatory lesions), ensemble strategies capable of leveraging these strengths are particularly valuable. The motivating application focuses on the classification of thyroid cancer nodules whose cytopathological clinical diagnosis is typically characterized by a high number of false positive cases that may result in unnecessary thyroidectomy. We apply and compare the performance of seven individual classifiers, along with four ensemble voting techniques (including Shapley values), in a real-world study focused on classifying thyroid cancer nodules using proteomic features obtained through mass spectrometry. Our results indicate a slight improvement in the classification accuracy for ensemble systems compared to the performance of single classifiers. Although the Shapley value-based voting method remains comparable to the other voting methods, we envision this new ensemble approach could be effective in improving the performance of single classifiers in further applications, especially when complementary algorithms are considered in the ensemble. The application of these techniques can lead to the development of new tools to assist clinicians in diagnosing thyroid cancer using proteomic features derived from mass spectrometry. Full article

Temporal domain generalization is crucial for the temporal forecasting of sensor data due to the non-stationary and evolving nature of most sensor-generated time series. However, temporal dynamics vary in scale, semantics, and structure, leading to distribution shifts that a single model cannot easily generalize over. Additionally, conflicts between temporal domain-specific patterns and limited model capacity make it difficult to learn shared parameters that work universally. To address this challenge, we propose an ensemble learning framework that leverages multiple domain-specific models to improve temporal domain generalization for sensor data forecasting. We first segment the original sensor time series into distinct temporal tasks to better handle the distribution shifts inherent in sensor measurements. A meta-learning strategy is then applied to extract shared representations across these tasks. Specifically, during meta-training, a recurrent encoder combined with variational inference captures contextual information for each task, which is used to generate task-specific model parameters. Relationships among tasks are modeled via a self-attention mechanism. For each query, the prediction results are adaptively reweighted based on all previously learned models. At inference, predictions are directly generated through the learned ensemble mechanism without additional tuning. Extensive experiments on public sensor datasets demonstrate that our method significantly enhances the generalization performance in forecasting across unseen sensor segments. Full article

Background: Poststroke dysphagia is a common condition that can lead to complications such as aspiration pneumonia and malnutrition, significantly affecting the quality of life. Most patients recover their swallowing function spontaneously, but in others difficulties persist beyond six months. Can we predict this in advance? On the other hand, there have been recent attempts to use machine learning to predict disease prognosis. Therefore, this study aims to investigate whether machine learning can predict the long-term prognosis for poststroke dysphagia using early videofluoroscopic swallowing study (VFSS) data. Methods: Data from VFSSs performed within 1 month of onset and swallowing status at 6 months were collected retrospectively in patients with dysphagia who experienced their first acute stroke at a university hospital. We selected 14 factors (lip closure, bolus formation, mastication, apraxia, tongue-to-palate contact, premature bolus loss, oral transit time, triggering of pharyngeal swallow, vallecular residue, laryngeal elevation, pyriform sinus residue, coating of the pharyngeal wall, pharyngeal transit time, and aspiration) from the VFSS data, scored them, and analyzed whether they could predict the long-term prognosis using five machine learning algorithms: Random forest, CatBoost classifier, K-neighbor classifier, Light gradient boosting machine, Extreme gradient boosting. These algorithms were combined through an ensemble method to create the final model. Results: In total, we collected data from 448 patients, of which 70% were used for training and 30% for testing. The final model was evaluated using accuracy, precision, recall, F1-score, and Area Under the Receiver Operating Characteristic Curve (AUC), resulting in values of 0.98, 0.94, 0.84, 0.88, and 0.99, respectively. Conclusions: Machine learning models using early VFSS data have shown high accuracy and predictive power in predicting the long-term prognosis of patients with poststroke dysphagia, and they are likely to provide useful information for clinicians. Full article

Generalising deep-learning models to perform well on unseen data domains with minimal retraining remains a significant challenge in computer vision. Even when the target task—such as quantifying the number of elements in an image—stays the same, data quality, shape, or form variations can deviate from the training conditions, often necessitating manual intervention. As a real-world industry problem, we aim to automate stock level estimation in retail cabinets. As technology advances, new cabinet models with varying shapes emerge alongside new camera types. This evolving scenario poses a substantial obstacle to deploying long-term, scalable solutions. To surmount the challenge of generalising to new cabinet models and cameras with minimal amounts of sample images, this research introduces a new solution. This paper proposes a novel ensemble model that combines DenseNet-201 and Vision Transformer (ViT-B/8) architectures to achieve generalisation in stock-level classification. The novelty aspect of our solution comes from the fact that we combine a transformer with a DenseNet model in order to capture both the local, hierarchical details and the long-range dependencies within the images, improving generalisation accuracy with less data. Key contributions include (i) a novel DenseNet-201 + ViT-B/8 feature-level fusion, (ii) an adaptation workflow that needs only two images per class, (iii) a balanced layer-unfreezing schedule, (iv) a publicly described domain-shift benchmark, and (v) a 47 pp accuracy gain over four standard few-shot baselines. Our approach leverages fine-tuning techniques to adapt two pre-trained models to the new retail cabinets (i.e., standing or horizontal) and camera types using only two images per class. Experimental results demonstrate that our method achieves high accuracy rates of 91% on new cabinets with the same camera and 89% on new cabinets with different cameras, significantly outperforming standard few-shot learning methods. Full article

Epilepsy remains a widespread neurological disorder, with approximately 30% of patients showing resistance to current antiepileptic therapies. To address this unmet need, a series of 2-(isoindolin-2-yl) esters derived from natural amino acids were designed and evaluated for their potential interaction with the GABA_A receptor. Sixteen derivatives were subjected to in silico assessments, including physicochemical and ADMET profiling, virtual screening–ensemble docking, and enhanced sampling molecular dynamics simulations (metadynamics calculations). Among these, compounds derived from the aromatic amino acids, phenylalanine, tyrosine, tryptophan, and histidine, exhibited superior predicted affinity, attributed to π–π stacking interactions at the benzodiazepine binding site of the GABA_A receptor. Based on computational performance, the tyrosine and tryptophan derivatives were synthesized and further assessed in vivo using the pentylenetetrazole-induced seizure model in zebrafish (Danio rerio). The tryptophan derivative produced comparable behavioral seizure reduction to the reference drug diazepam at the tested concentrations. The results implies that aromatic amino acid-derived isoindoline esters are promising anticonvulsant candidates and support the hypothesis that π–π interactions may play a critical role in modulating GABA_A receptor binding affinity. Full article

Background/Objectives: Accurate classification of brain tumors is critical for treatment planning and prognosis. While deep convolutional neural networks (CNNs) have shown promise in medical imaging, few studies have systematically compared multiple architectures or integrated ensemble strategies to improve diagnostic performance. This study aimed to evaluate various CNN models and optimize classification performance using a majority voting ensemble approach on T1-weighted MRI brain images. Methods: Seven pretrained CNN architectures were fine-tuned to classify four categories: glioblastoma, meningioma, pituitary adenoma, and no tumor. Each model was trained using two optimizers (SGDM and ADAM) and evaluated on a public dataset split into training (70%), validation (10%), and testing (20%) subsets, and further validated on an independent external dataset to assess generalizability. A majority voting ensemble was constructed by aggregating predictions from all 14 trained models. Performance was assessed using accuracy, Kappa coefficient, true positive rate, precision, confusion matrix, and ROC curves. Results: Among individual models, GoogLeNet and Inception-v3 with ADAM achieved the highest classification accuracy (0.987). However, the ensemble approach outperformed all standalone models, achieving an accuracy of 0.998, a Kappa coefficient of 0.997, and AUC values above 0.997 for all tumor classes. The ensemble demonstrated improved sensitivity, precision, and overall robustness. Conclusions: The majority voting ensemble of diverse CNN architectures significantly enhanced the performance of MRI-based brain tumor classification, surpassing that of any single model. These findings underscore the value of model diversity and ensemble learning in building reliable AI-driven diagnostic tools for neuro-oncology. Full article

Breast cancer is a leading cause of mortality among women worldwide. Early detection is crucial for increasing patient survival rates. Artificial intelligence, particularly convolutional neural networks (CNNs), has enabled the development of effective diagnostic systems by digitally processing mammograms. CNNs have been widely used for the classification of breast cancer in images, obtaining accurate results similar in many cases to those of medical specialists. This work presents a hybrid feature extraction approach for breast cancer detection that employs variants of EfficientNetV2 network and convenient image representation based on phase features. First, a region of interest (ROI) is extracted from the mammogram. Next, a three-channel image is created using the local phase, amplitude, and orientation features of the ROI. A feature vector is constructed for the processed mammogram using the developed CNN model. The size of the feature vector is reduced using simple statistics, achieving a redundancy suppression of $99.65\%$. The reduced feature vector is classified as either malignant or benign using a classifier ensemble. Experimental results using a training/testing ratio of 70/30 on 15,506 mammography images from three datasets produced an accuracy of $86.28\%$, a precision of $78.75\%$, a recall of $86.14\%$, and an F1-score of $80.09\%$ with the modified EfficientNetV2 model and stacking classifier. However, an accuracy of $93.47\%$, a precision of $87.61\%$, a recall of $93.19\%$, and an F1-score of $90.32\%$ were obtained using only CSAW-M dataset images. Full article

Fire spread between buildings is the primary cause of extensive fire damage in traditional village timber structure clusters. Accurately assessing fire spread risk is crucial for the preservation of these architectural ensembles. During the development and conservation of traditional villages, fire risk dynamics may shift due to fire-resistant retrofits or layout modifications, necessitating repeated risk reevaluations. To address challenges such as the computational intensity of fire spread simulations, high costs, and data acquisition difficulties, this study proposes a directed graph-based method for fire spread risk analysis and risk level classification in timber structure clusters, accounting for their unique fire propagation characteristics. First, localized fire spread paths and propagation times between nodes (buildings) are determined through fire spread simulations, constructing an adjacency matrix for the directed graph of the building cluster. Path search algorithms then identify the spread range and velocity under specific fire scenarios. Subsequently, a zoned risk assessment model for individual buildings is developed based on critical fire spread loss and velocity, integrating each building’s fire resistance and its probability of exposure to different risk zones to determine the overall cluster’s fire spread risk level. The method is validated using a case study of a typical village in Yunnan Province. Results demonstrate that the approach efficiently computes fire spread characteristics across different scenarios and quantitatively evaluates risk levels, enabling targeted fire safety interventions based on village-specific spread patterns. Case analysis reveals significant variations in fire spread behavior: Village 1, Village 2, and Village 3 exhibit fire resistance indices of 0.59, 0.757, and 0.493, corresponding to high, moderate, and high fire spread risk levels, respectively. Full article

Accurate estimation of long-term average (LTA) discharge is fundamental for water resource assessment, infrastructure planning, and hydrological modeling, yet it remains a significant challenge, particularly in data-scarce or ungauged basins. This study introduces an advanced machine learning framework to estimate long-term average discharge using globally available hydrological station metadata from the Global Runoff Data Centre (GRDC). The methodology involved comprehensive data preprocessing, extensive feature engineering, log-transformation of the target variable, and the development of multiple predictive models, including a custom deep neural network with specialized pathways and gradient boosting machines (XGBoost, LightGBM, CatBoost). Hyperparameters were optimized using Bayesian techniques, and a weighted Meta Ensemble model, which combines predictions from the best individual models, was implemented. Performance was rigorously evaluated using R², RMSE, and MAE on an independent test set. The Meta Ensemble model demonstrated superior performance, achieving a Coefficient of Determination (R²) of 0.954 on the test data, significantly surpassing baseline and individual advanced models. Model interpretability analysis using SHAP (Shapley Additive explanations) confirmed that catchment area and geographical attributes are the most dominant predictors. The resulting model provides a robust, accurate, and scalable data-driven solution for estimating long-term average discharge, enhancing water resource assessment capabilities and offering a powerful tool for large-scale hydrological analysis. Full article

Despite advances in automation, the global manufacturing sector continues to rely heavily on manual package inspection, creating bottlenecks in production and increasing labor demands. Although disruptive technologies such as big data analytics, smart sensors, and machine learning have revolutionized industrial connectivity and strategic decision-making, real-time quality control (QC) on conveyor lines remains predominantly analog. This study proposes an intelligent package integrity monitoring system that integrates waste reduction strategies with both narrow and Generative AI approaches. Narrow AI models were deployed to detect package damage at full line speed, aiming to minimize manual intervention and reduce waste. Using a synthetically generated dataset of 200 paired top-and-side package images, we developed and evaluated 10 distinct detection pipelines combining various algorithms, image enhancements, model architectures, and data processing strategies. Several pipeline variants demonstrated high accuracy, precision, and recall, particularly those utilizing a YOLO v8 segmentation model. Notably, targeted preprocessing increased top-view MobileNetV2 accuracy from chance to 67.5%, advanced feature extractors with full enhancements achieved 77.5%, and a segmentation-based ensemble with feature extraction and binary classification reached 92.5% accuracy. These results underscore the feasibility of deploying AI-driven, real-time QC systems for sustainable and efficient manufacturing operations. Full article

Predicting workload for tasks in manufacturing is a complex challenge due to the numerous variables involved. In small- and medium-sized enterprises (SMEs), this process is often experience-based, leading to inaccurate predictions that significantly impact production planning, order management, and consequently the ability to meet customer deadlines. This paper presents an approach that leverages machine learning to enhance workload prediction with minimal data collection, making it particularly suitable for SMEs. A case study application using supervised machine learning models for regression, trained in an open-source data analytics, reporting, and integration platform (KNIME Analytics Platform), has been carried out. An Automated Machine Learning (AutoML) regression approach was employed to identify the most suitable model for task workload prediction based on minimising the Mean Absolute Error (MAE) scores. Specifically, the Regression Tree (RT) model demonstrated superior accuracy compared to more traditional simple averaging and manual predictions when modelling data for a single product type. When incorporating all available product data, despite a slight performance decrease, the XGBoost Tree Ensemble still outperformed the traditional approaches. These findings highlight the potential of machine learning to improve workload forecasting in manufacturing, offering a practical and easily implementable solution for SMEs. Full article

Cities house more than half of the world’s population and are responsible for more than 70% of the world anthropogenic CO₂ emissions. Therefore, quantifications of emissions from major cities, which are only less than a hundred intense emitting spots across the globe, should allow us to monitor changes in global fossil fuel CO₂ emissions in an independent, objective way. The study adopted a high-spatiotemporal-resolution regional assimilation method using satellite observation data and atmospheric transport model WRF-Chem/DART to assimilate CO₂ concentration and fluxes in Lisbon, a major city in Portugal. It is based on Zhang’s assimilation method, combined OCO-2 XCO₂ retrieval data, ODIAC 1 km anthropogenic CO₂ emissions and Ensemble Adjustment Kalman Filter Assimilation. By employing three two-way nested domains in WRF-Chem, we refined the spatial resolution of the CO₂ concentrations and fluxes over Lisbon to 3 km. The spatiotemporal distribution characteristics and main driving factors of CO₂ concentrations and fluxes in Lisbon and its surrounding cities and countries were analyzed in March 2020, during the period affected by COVID-19 pandemic. The results showed that the monthly average CO₂ and XCO₂ concentrations in Lisbon were 420.66 ppm and 413.88 ppm, respectively, and the total flux was 0.50 Tg CO₂. From a wider perspective, the findings provide a scientific foundation for urban carbon emission management and policy-making. Full article

This study aimed to investigate population genetic differences related to reproductive traits between Duroc and Yorkshire (Dutch Large White) pigs using two approaches: linear mixed models that dissect additive and dominant effects, and selective sweep analysis. (1) Methods: Genome-wide single-nucleotide polymorphism (SNP) data of 3917 Duroc and 3217 Yorkshire pigs were analyzed. The first principal component (PC1) was used as a simulated phenotype to capture population-level variance. Additive and dominant genetic effects were partitioned and evaluated by using the combination of the linear mixed models (LMM) and ADDO’s algorithm (LMM + ADDO). In parallel, selective sweep signals were detected using fixation index (F_ST) and nucleotide diversity (θ_π) analyses. A comparative assessment was then conducted between the LMM + ADDO and the selective sweep analysis results. Significant loci were annotated using quantitative trait loci (QTL) databases and the Ensembl genome browser. (2) Results: There are 39040 SNPs retained after quality control. Using the LMM + ADDO framework with PC1 as a simulated phenotype, a total of 632 significant SNPs and 184 candidate genes were identified. Notably, 587 SNPs and 171 genes were uniquely detected by the LMM + ADDO method and not among loci detected by the top 5% of F_ST and θ_π values. Key candidate genes associated with litter size included HSPG2, KAT6B, SAMD8, and LRMDA, while DLGAP1, MYOM1, and VTI1A were associated with teat number traits. (3) Conclusions: This study demonstrates the power of integrating additive and dominant effect modeling with population genetics approaches for the detection of genomic regions under selection. The findings provide novel insights into the genetic architecture of reproductive traits in pigs and have practical implications for understanding the inheritance of complex traits. Full article