Search Results (1,504)

Lipschitz-based classification provides a flexible framework for general metric spaces, naturally adapting to complex data structures without assuming linearity. However, direct applications of classical extensions often yield decision boundaries equivalent to the 1-Nearest Neighbour classifier, leading to overfitting and sensitivity to noise. Addressing this limitation, this paper introduces a novel binary classification algorithm that integrates probabilistic kernel smoothing with explicit Lipschitz extensions. We approximate the conditional probability of class membership by extending smoothed labels through a family of bounded Lipschitz functions. Theoretically, we prove that while direct extensions of binary labels collapse to nearest-neighbour rules, our probabilistic approach guarantees controlled complexity and stability. Experimentally, evaluations on synthetic and real-world datasets demonstrate that this methodology generates smooth, interpretable decision boundaries resilient to outliers. The results confirm that combining kernel smoothing with adaptive Lipschitz extensions yields performance competitive with state-of-the-art methods while offering superior geometric interpretability. Full article

Background/Objectives: Vision–language models such as BiomedCLIP are increasingly investigated for their diagnostic potential in medical imaging. Although these foundation models show promise in general radiographic interpretation, their application in pediatric domains—particularly for subtle, postoperative findings like esophageal strictures—remains underexplored. This study aimed to evaluate the diagnostic performance of BiomedCLIP in classifying pediatric esophageal radiographs into three clinically relevant categories: presence of contrast agent, full esophageal visibility, and presence of esophageal stricture. Methods: We retrospectively analyzed 143 pediatric esophageal X-rays collected between 2021 and 2025. Each image was annotated by two pediatric radiology experts and categorized according to esophageal visibility, contrast presence, and stricture occurrence. BiomedCLIP was used in a zero-shot classification setup without fine-tuning. Model predictions were converted into binary outcomes and assessed against the ground truth using a comprehensive suite of 27 performance metrics, including accuracy, sensitivity, specificity, F1-score, AUC, and calibration analyses. Results: BiomedCLIP achieved high precision (88.7%) and a favorable AUC (85.4%) in detecting contrast agent presence, though specificity remained low (20%), leading to a high false-positive rate. The model correctly identified all cases of non-visible esophagus, but was untestable in predicting full visibility due to the absence of positive cases. Critically, its performance in detecting esophageal strictures was poor, with accuracy at 24%, sensitivity at 44%, specificity at 18%, and AUC of 0.26. Statistical overlap between contrast and stricture predictions indicated a lack of semantic differentiation within the model’s latent space. Conclusions: BiomedCLIP shows potential in detecting high-salience features such as contrast but fails to reliably identify esophageal strictures. Limitations include class imbalance, absence of fine-tuning, and architectural constraints in recognizing subtle morphologic abnormalities. These findings emphasize the need for domain-specific adaptation of foundation models before clinical implementation in pediatric radiology. Full article

Despite the growing use of machine learning (ML) for analyzing service quality and customer satisfaction, empirical studies based on Brazilian broadband telecommunications data remain scarce. This is especially true for those who leverage publicly available nationwide datasets. To address this gap, this study investigates customer satisfaction with broadband internet services in Brazil using supervised ML and explainable artificial intelligence (XAI) techniques applied to survey data collected by ANATEL between 2017 and 2020. Customer satisfaction was operationalized using the Net Promoter Score (NPS) reference scale, and three modifications in the scale were evaluated: (i) a binary model grouping ratings ≥ 8 as satisfied and ≤7 as dissatisfied (portion of the neutrals as satisfied and another as dissatisfied); (ii) a binary model excluding neutral responses (ratings 7–8) and retaining only detractors (≤6) and promoters (≥9); and (iii) a multiclass model following the original NPS categories (detractors, neutrals, and promoters). Nine ML classifiers were trained and validated on tabular data for each formulation. Model interpretability was addressed through SHAP and feature importance analysis using tree-based models. The results indicate that Histogram Gradient Boosting and Random Forest achieve the most robust and stable performance, particularly in binary classification scenarios. The analysis of neutral customers reveals classification ambiguity, showing scores of “7” tend toward dissatisfaction, while scores of “8” tend toward satisfaction. XAI analyses consistently identify browsing speed, billing accuracy, fulfillment of advertised service conditions, and connection stability as the most influential predictors of satisfaction. By combining predictive performance with model transparency, this study provides computational evidence for explainable satisfaction modeling and highlights the value of public regulatory datasets for reproducible ML research. Full article

Background/Objectives: Obstructive sleep apnea (OSA) is a prevalent, yet underdiagnosed, disorder associated with cardiovascular and cognitive risks. While overnight polysomnography (PSG) remains the diagnostic gold standard, it is resource-intensive and impractical for large-scale rapid screening. Methods: This study extends prior work on feature extraction and binary classification using tracheal breathing sounds (TBS) and anthropometric data by introducing a meta-modeling framework that utilizes machine learning (ML) and aggregates six one-vs.-one classifiers for multi-class OSA severity prediction. We employed out-of-bag (OOB) estimation and three-fold cross-validation to assess model generalization performance. To enhance reliability, the framework incorporates conformal prediction to provide calibrated confidence sets. Results: In the three-class setting (non, mild, moderate/severe), the model achieved 76.7% test accuracy, 77.7% sensitivity, and 87.1% specificity, with strong OOB performance of 91.1% accuracy, 91.6% sensitivity, and 95.3% specificity. Three-fold confirmed stable performance across folds (mean accuracy: 77.8%; mean sensitivity: 78.6%; mean specificity: 76.4%) and conformal prediction achieved full coverage with an average set size of 2. In the four-class setting (non, mild, moderate, severe), the model achieved 76.7% test accuracy, 75% sensitivity, and 92% specificity, with OOB performance of 88.2% accuracy, 91.6% sensitivity, and 88.2% specificity. Conclusions: These findings support the potential of this non-invasive system as an efficient and rapid OSA severity assessment whilst awake, offering a scalable alternative to PSG for large-scale screening and clinical triaging. Full article

This study presents an edge-based intrusion detection methodology designed to enhance cybersecurity in Internet of Things environments, which remain highly vulnerable to complex attacks. The approach employs an Auxiliary Classifier Generative Adversarial Network capable of classifying network traffic in real-time while simultaneously generating high-fidelity synthetic data within a unified framework. The model is implemented in TensorFlow and deployed on the energy-efficient NVIDIA Jetson Orin Nano, demonstrating the feasibility of executing advanced deep learning models at the edge. Training is conducted on network traffic collected from diverse IoT devices, with preprocessing focused on TCP-based threats. The integration of an auxiliary classifier enables the generation of labeled synthetic samples that mitigate data scarcity and improve supervised learning under imbalanced conditions. Experimental results demonstrate strong detection performance, achieving a precision of 0.89 and a recall of 0.97 using the standard 0.5 decision threshold inherent to the sigmoid-based binary classifier, indicating an effective balance between intrusion detection capability and false-positive reduction, which is critical for reliable operation in IoT scenarios. The generative component enhances data augmentation, robustness, and generalization. These results show that combining generative adversarial learning with edge computing provides a scalable and effective approach for IoT security. Future work will focus on stabilizing training procedures and refining hyperparameters to improve detection performance while maintaining high precision. Full article

This study investigates the relationship between religiosity and attitudes toward COVID-19 vaccination in Romania using nationally representative survey data from the Barometer of Religious Life (December 2021). Five survey items measuring religious beliefs and practices were used to construct a Religious Practice Index, whose reliability and one-dimensionality were confirmed through Cronbach’s Alpha and factor analysis. Correlation analysis revealed a small but statistically significant negative association between religiosity and vaccination acceptance (r = −0.106, p = 0.001). Binary logistic regression further indicated that higher religiosity, younger age, lower income, and rural residence were significant predictors of reduced vaccination likelihood, while older age, higher income, and urban residence were associated with greater acceptance. Nevertheless, the model explained only 9.3% of the variance and correctly classified 64.4% of cases, suggesting modest predictive power. These findings indicate that religiosity influences vaccination attitudes but does not serve as a dominant predictor, highlighting the importance of other additional factors that were beyond the scope of this analysis and were not measured. Full article

Refill Friction Stir Spot Welding (RFSSW) provides significant advantages over competing spot joining technologies, but detecting RFSSW’s often small and subtle defects remains challenging. In this study, kinematic feedback data from a RFSSW machine’s factory-installed sensors was used to successfully predict defect presence with 96% accuracy (F1 = 0.92) and preliminary multi-class defect diagnosis with 84% accuracy (F1 = 0.82). Thirty adverse treatments (e.g., contaminated coupons, worn tools, and incorrect material thickness) were carried out to create 300 potentially defective welds, plus control welds, which were then evaluated using profilometry, computed tomography (CT) scanning, cutting and polishing, and tensile testing. Various machine learning (ML) models were trained and compared on statistical features, with support vector machine (SVM) achieving top performance on final quality prediction (binary), random forest outperforming other models in classifying welds into six diagnosis categories (plus a control category) based on the adverse treatments. Key predictors linking process signals to defect formation were identified, such as minimum spindle torque during the plunge phase. In conclusion a framework is proposed to integrate these models into a manufacturing setting for low-cost, full-coverage evaluation of RFSSWs. Full article

Background: Low and middle-income countries (LMICs) such as Mexico are experiencing rapid population aging, accompanied by increasing levels of functional dependency and growing long-term care (LTC) needs. Objectives: We aimed to identify the factors associated with varying levels of functional dependency in order to assist population health planning and LTC policy in aging populations in Mexico. Methods: This cross-sectional study analyzed data from the 2021 wave of the Mexican Health and Aging Study (MHAS). Functional dependency was assessed through a modified Autonomie Gérontologie Groupes Iso-Ressources (AGGIR) scale, adapted to incorporate cognitive and physical assessments suitable for the Mexican context. Socioeconomic, health-related, and psychological variables were examined using ordinal logistic regression models. Results: Among 8049 participants included in the analysis, 87.08% were classified with non-to-mild dependency, 9.13% with moderate dependency, and 3.79% with severe dependency. More severe levels of functional dependency were associated with older age, lower educational attainment, not having a partner (being single, widowed, separated or divorced), and the presence of chronic conditions such as hypertension and cardiovascular disease. Conclusions: In contrast, higher educational attainment and regular physical activity were associated with less severe levels of dependency. These associations highlight the multifactorial nature of dependency in later life. The application of a graded, multidimensional dependency classification provides a more comprehensive and differentiated understanding of care needs than binary functional measures. This population-level perspective may support the prioritization of healthy aging strategies and long-term care planning in rapidly aging middle-income settings such as Mexico. Full article

This study addresses an interpretable supervised binary classification problem under constrained feature availability and class imbalance. The objective is to evaluate whether reliable predictive performance can be achieved using exclusively pre-event administrative variables while preserving transparency and analytical traceability of model decisions. A comparative framework is developed using linear and ensemble-based classifiers, combined with resampling strategies and exhaustive hyperparameter optimization embedded within cross-validation. Model performance is evaluated using standard classification metrics, with particular emphasis on the Matthews correlation coefficient as a robust measure under imbalance. In addition to predictive accuracy, the analysis incorporates global, structural, and local interpretability mechanisms, including permutation feature importance, explicit decision paths derived from tree-based models, and additive local explanations. Experimental results show that optimized ensemble models achieve consistent performance gains over linear baselines while maintaining a balanced error structure across classes. Importantly, the most influential predictors exhibit stable rankings across models and explanation methods, indicating a concentrated and robust discriminative signal within the constrained feature space. The interpretability analysis demonstrates that complex classifiers can be decomposed into verifiable decision rules and locally coherent feature contributions. Overall, the findings confirm that interpretable supervised classification can be reliably conducted under administrative data constraints, providing a reproducible modeling framework that balances predictive performance, error analysis, and explainability in applied mathematical settings. Full article

In many land cover classification tasks, the limited precision of individual sensors hinders the accurate separation of certain classes, largely due to the complexity of the Earth’s surface morphology. To mitigate these issues, decision fusion methodologies are employed, allowing data from multiple sensors to be synthesized into robust and more conclusive classification outcomes. This study employs fully polarimetric Synthetic Aperture Radar (PolSAR) imagery and leverages the strengths of three decomposition methods, namely Pauli’s, Krogager’s, and Cloude’s, by extracting their respective components for improved detection. From each decomposition method, three scattering components are derived, enabling the extraction of informative features that describe the scattering behavior associated with various land cover types. The extracted scattering features, treated as independent sensors, were used to train three neural network classifiers. The resulting outputs were then considered as local decisions for each land cover type and subsequently fused through a decision fusion rule to generate more complete and accurate classification results. Experimental results demonstrate that the proposed Multi-Class Bahadur–Lazarsfeld Expansion (MC-BLE) fusion significantly enhances classification performance, achieving an overall accuracy (OA) of 95.78% and a Kappa coefficient of 0.94. Compared to individual classification methods, the fusion notably improved per-class accuracy, particularly for complex land cover boundaries. The core innovation of this work is the transformation of the Bahadur–Lazarsfeld Expansion (BLE), originally designed for binary decision fusion into a multi-class framework capable of addressing multiple land cover types, resulting in a more effective and reliable decision fusion strategy. Full article

The collision dynamics of binary seawater droplets are pivotal in marine engineering applications, like spray desalination and engine cooling. While high-fidelity simulations can resolve these dynamics, they are computationally prohibitive for rapid design and analysis. This study introduces the first interpretable machine learning (ML) framework to predict and elucidate the collision outcomes of head-on binary seawater droplets. A high-fidelity numerical dataset, generated via Modified Coupled Level Set-VOF (M-CLSVOF) simulations across a broad Weber number (We) range, serves as the foundation for training multiple classifiers. Among the tested algorithms, the Random Forest model achieved superior performance with 96.2% accuracy. The model’s predictions precisely identified the critical Weber number for the transition from coalescence to reflexive separation at We ≈ 22.3 for seawater. Moving beyond black-box prediction, we employed SHapley Additive exPlanations (SHAP) to quantitatively deconstruct the model’s decision-making process. SHAP analysis confirmed the dominance of the Weber number (75% contribution) and revealed the context-dependent role of the Reynolds number (25% contribution) in modulating the collision outcome. Furthermore, a comparative analysis with freshwater droplets quantified a 6% elevation in the critical Weber number for seawater, attributed to salinity-induced modifications in fluid properties. Finally, a machine-learned regime map in the We-Ohnesorge space was constructed, delineating the coalescence and separation boundaries. This work establishes ML as a powerful, interpretable surrogate model that not only delivers rapid, accurate predictions but also extracts fundamental physical insights, offering a valuable paradigm for optimizing marine spray systems. Full article

Condition monitoring and diagnosis in a permanent magnet synchronous generator (PMSG) are crucial for ensuring its service continuity and reliability. Recent advancements have introduced innovative, non-invasive techniques for detecting mechanical and electrical faults in this machine. This paper proposes a novel application of the discrete wavelet transform and binary classifiers for diagnosing interturn short-circuit faults in a PMSG with high accuracy and low computational burden. The objective of fault diagnosis is to detect the presence of an interturn short-circuit fault (fault vs. no-fault) under different fault severities and operating speeds. Multiple binary models were trained separately for each fault scenario. The three-phase currents from the PMSG are processed using the discrete wavelet transform to extract features, which are then fed into a binary classifier based on a Random Forest algorithm. Optimization techniques are used to improve the performance of the binary classifiers. Experimental results obtained under various stator fault conditions in the PMSG are presented. Metrics such as accuracy and confusion matrices are used to evaluate the performance of binary classifiers. Full article

The rapid adoption of Machine Learning (ML) in high-impact domains has intensified the need for systematic tools to assess and improve the trustworthiness of predictive models beyond conventional performance metrics. This paper presents TWEEF (Trustworthiness Estimation and Enhancement Framework), a modular and extensible framework that operationalizes trustworthiness through the joint evaluation of performance, fairness, and interpretability. TWEEF integrates intuitionistic fuzzy logic and subjective logic to transform quantitative trust-related metrics into linguistic assessments, which are subsequently aggregated using operators such as the Linguistic Weighted Average (LWA), Gaussian Weighted Aggregation (GWA), and Subjective Logic (SL). The framework extends the scikit-learn ecosystem through a meta-estimator, the TrustworthyClassifier, which orchestrates metric computation, bias-mitigation procedures, surrogate-model generation, and trust aggregation within a unified, pipeline-compatible workflow. The framework is empirically evaluated through four experiments on widely used benchmark datasets (German Credit, COMPAS, and Adult) in binary classification settings. Results show that TWEEF consistently reveals fairness and interpretability limitations that may remain hidden when relying solely on predictive performance, and that the resulting trust scores respond coherently to different metric configurations and weighting schemes. These findings indicate that TWEEF provides a structured mechanism for trust assessment and enhancement, while also offering a flexible foundation for future extensions to additional learning tasks and evaluation dimensions. Full article

Classifying optical water types (OWTs), particularly concerning different phytoplankton bloom types, is critically important because dominant phytoplankton groups govern key marine ecosystem functions and biogeochemical processes, including nutrient cycling and carbon export. This study refines a recent OWT classification method developed for the Second-Generation Global Imager (SGLI), which was originally proposed to discriminate dinoflagellate and diatom blooms. By employing binary logistic regression (bLR) with independent in situ data from Karenia selliformis (dinoflagellate) blooms off the Kamchatka Peninsula and Skeletonema spp. (diatom) blooms in Tokyo Bay, this study establishes more robust and statistically meaningful boundaries between OWTs. The analysis confirms the diagnostic spectral shapes from SGLI data: a trough at 490 nm for K. selliformis blooms and a peak at 490 nm for diatom blooms, validating the consistency of this spectral criterion. The updated method reliably identifies waters dominated by coloured dissolved organic matter and different phytoplankton functional types in mesotrophic waters, and successfully detected a Karenia mikimotoi bloom in the Gulf St. Vincent, South Australia, demonstrating its potential for the global monitoring of red tides. By providing a reliable, satellite-based tool to distinguish between ecologically distinct phytoplankton groups, this refined OWT classification offers a valuable data product to improve the accuracy of marine ecosystem and carbon cycle models, moving beyond bulk chlorophyll-a parameterizations. Full article

Glioblastoma multiforme (GBM) is the most aggressive primary brain cancer (with a median survival time of 14.5 months), characterized by heterogeneity. Identifying prognostic molecular subtypes could provide a deeper exposition of GBM biology with potential therapeutic implications. In this study, we classified GBM into two prognostic subtypes, C1-GBM (n = 57; OS: 313 days) and C2-GBM (n = 109; OS: 452 days), using pathway-based signatures derived from RNA-seq data. Unsupervised consensus clustering revealed that only binary classification (cluster number, CN = 2; mean cluster consensus score = 0.84) demonstrated statistically prognostic differences. We characterized C1 and C2 based on oncogenic pathway and immune signatures. Specifically, C1-GBM was categorized as an immune-infiltrated “hot” tumor, with high infiltration of immune cells, particularly macrophages and CD4⁺ T cells, while C2-GBM as an “inherent driving” subtype, showing elevated activity in G2/M checkpoint genes. To predict the C1 or C2 classification and explore therapeutic interventions, we developed a neural network model. By using Weighted Correlation Network Analysis (WGCNA), we obtained the gene co-expression module based on both gene expression pattern and distribution among patients in TCGA dataset (n = 166) and identified nine hub genes as potentially prognostic biomarkers for the neural network. The model showed strong accuracy in predicting C1/C2 classification and prognosis, validated by the external CGGA-GBM dataset (n = 85). Based on the classification of the BP neural network model, we constructed a Cox nomogram prognostic prediction model for the TCGA-GBM dataset. We predicted potential therapeutic small molecular drugs by targeting subtype-specific oncogenic pathways and validated drug sensitivity (C1-GBM: Methotrexate and Cisplatin; C2-GBM: Cytarabine) by assessing IC₅₀ values against GBM cell lines (divided into C1/C2 subtypes based on the nine hub genes) from the Genomics of Drug Sensitivity in Cancer database. This study introduces a pathway-based prognostic molecular classification of GBM with “hot” (C1-GBM) and “inherent driving” (C2-GBM) tumor subtypes, providing a prediction model based on hub biomarkers and potential therapeutic targets for treatments. Full article