Search Results (9,906)

Rapid growth in connected devices, artificial intelligence applications, and the Internet of Things (IoT) is driving demand for ultra-high data rates, low latency, and energy-efficient communication infrastructure. Wireless optical communication (WOC), including free-space optical (FSO), is recognized as a disruptive technology for 6G and future-generation networks. However, atmospheric visibility critically affects WOC/FSO link availability, capacity, and reliability. This study proposes a machine learning (ML)-based low-visibility classification model that integrates Meteorological Aerodrome Reports (METARs) with microwave-link received-signal (Rx) features. Visibility below 6000 m is predicted at the 1 h, 3 h, and 6 h horizons using 18 months of data from Suvarnabhumi Airport (VTBS) and Don Mueang Airport (VTBD) in Bangkok, Thailand. Four ML algorithms, namely logistic regression, random forest, extreme gradient boosting, and light gradient boosting machine (LGBM), are evaluated against persistence and Terminal Aerodrome Forecast (TAF) baselines. In a 100-round block-bootstrap evaluation, LGBM with METAR-Rx achieved the highest mean F1 scores at the 1 h and 3 h horizons, outperforming TAF by 28 and 20 percentage points at the 1 h horizon for VTBS and VTBD, respectively. SHAP and ablation analyses suggested that current visibility is the dominant predictor, while Rx features provide complementary information and improve F1 performance by approximately 1–4 percentage points. Seasonal analysis shows stronger cool-season performance, while rainy-season prediction remains challenging. Adding visibility-trend features further improves performance, with the best combined model achieving 1 h F1 scores of 0.7253 for VTBS and 0.6495 for VTBD. These findings indicate that integrating the METAR-Rx feature set can support short-term low-visibility classification. Full article

This paper presents a numerical-computational analysis of ${\mathsf{\ell }}_2$-regularized logistic learning for binary intrusion detection under heterogeneous datasets, class imbalance, and cross-dataset shift. Rather than proposing a new intrusion detection architecture, the study examines how numerical conditioning, feature scaling, feature set design, threshold selection, false negative behavior, false alarm behavior, and distribution shift affect operational detection performance. Experiments were conducted on CICIDS2017, UNSW-NB15, and CIRA-CIC-DoHBrw-2020 using reproducible train–validation–test protocols over five fixed random seeds. The numerical audit showed that standard scaling reduced the spectral condition number of traffic feature matrices by several orders of magnitude across datasets and feature configurations. However, scaling did not produce uniformly monotonic predictive gains: in some cases, raw feature optimization achieved comparable or higher F1-score, whereas scaled preprocessing produced more controlled false alarm behavior. In-domain experiments showed that dataset-specific features may improve ranking metrics such as area under the receiver-operating-characteristic curve (AUROC) or area under the precision–recall curve (AUPR) without necessarily improving thresholded operational metrics. Cross-dataset transfer experiments revealed strong source–target asymmetry, with transferred thresholds producing either near-zero positive detection or excessive false alarms. Additional robustness experiments with Random Forest and XGBoost improved in-domain F1-score and false negative rate (FNR), but did not eliminate off-domain degradation, with high FNR persisting under direct cross-dataset transfer. Finally, a Kolmogorov–Smirnov-based distribution shift analysis showed that in-domain discrepancies were small, whereas cross-dataset discrepancies were consistently large under common standardized traffic features. These findings suggest that numerical stability, ranking quality, thresholded detection performance, false negative and false alarm behavior, and distribution shift should be analyzed jointly when evaluating intrusion detection models. Full article

Accurate correction of gridded numerical weather prediction precipitation forecasts remains challenging because many end-to-end deep learning correction models treat meteorological variables as undifferentiated data channels and therefore provide limited physical interpretability. Inspired by general principles of biological environmental sensing, selective information processing, and regulatory constraint learning, this study proposes PCPNet, a bio-inspired and physically constrained precipitation correction framework. The framework does not imitate a specific biological organ or species; instead, it abstracts three information-processing principles into a meteorological correction task. First, key weather-system cues, including low-level shear lines, trough-ridge effects, upper-level jet-stream forcing, vorticity-divergence-related vertical motion, and water-vapor flux convergence, are quantified as structured diagnostic fields. This transforms the subjective synoptic diagnosis of forecasters into automated grid-based sensing features. Second, these diagnostic cues are fused with numerical weather prediction variables and terrain descriptors in an encoder–attention–decoder network, allowing the model to emphasize dynamically important precipitation-triggering regions. Third, water-vapor conservation and terrain-forcing relationships are embedded as differentiable constraint losses, providing training-time constraint-based regulation that guides the corrected precipitation field toward physically consistent solutions. The method is evaluated from 2021 to 2023 in Hunan Province, China, using hourly numerical weather prediction model outputs as input features, China Meteorological Administration Land Data Assimilation System gridded analysis data as the training target, and independent meteorological station observations for strict cross-validation. PCPNet reduces the mean absolute error by 22.1% compared with the uncorrected China Meteorological Administration Land Data Assimilation System gridded precipitation products and outperforms Linear Regression, Bagging, Boosting, Multi-Layer Perceptron, TabNet, and Tree-based Progressive Regression Models by 12.9%, 13.5%, 16.9%, 10.8%, 14.9%, and 15.9%, respectively. The single-day event analysis provides an initial demonstration of heavy precipitation recovery capability, while comprehensive validation across long-term continuous weather events is planned for future operational deployment to further verify model stability. These results indicate that bio-inspired sensing and regulatory constraint learning can improve both the accuracy and interpretability of precipitation nowcasting correction. Full article

Traditional methods for determining starch content in corn kernels are labor-intensive, destructive, and inefficient. To overcome these challenges, this work developed a rapid, non-destructive approach based on near-infrared hyperspectral imaging, applied to 58 rainfed corn varieties. A spectral preprocessing scheme combining wavelet transform, multiplicative scatter correction, and standard normal variate transformation was employed to enhance spectral quality. A two-stage wavelength selection framework was established using competitive adaptive reweighted sampling and sparrow search algorithm optimization. From the selected optimal wavelengths, four predictive models, namely partial least squares regression, artificial neural network (ANN), convolutional neural networks, and gradient boosting decision tree, were established, implemented, and systematically compared. The results identify 14 key wavelengths (1020.65–1647.71 nm) strongly correlated with starch content, with clear assignments to specific chemical bonds and good physical interpretability. Among these models, the ANN exhibited the best performance. The R², RMSE, and RPD of the test set were 0.826, 0.759%, and 2.40, respectively, indicating favorable prediction accuracy and generalization ability. These key wavelengths provide a foundation for developing portable detection instruments. This work supports corn quality grading, breeding of high-starch varieties, and rapid raw material screening, thereby enhancing the quality and efficiency of the corn industry. Full article

This study proposes a sequential Hybrid GP-ANN architecture based on the concept of autonomous feature engineering for multi-output performance prediction of single-phase induction motors, serving as a high-fidelity surrogate model ready for digital twin integration. The required high-resolution motor dataset (88,375 samples) was generated using a parametric sweep approach with Ansys RMxprt. In the proposed architecture, Genetic Programming (GP) is not positioned as a final predictor but as an analytical filter that discovers hidden physical relationships in raw input data and converts them into super features, thereby structurally eliminating the polynomial explosion risk and scale sensitivity that arise when raw data are modelled directly. The nonlinear physical relationships discovered autonomously by GP are added to the network input matrix as mathematical vectors, breaking the black-box structure of standard artificial neural networks and transforming it into a physics-inspired grey-box model. The nonlinear terms discovered by GP are fed into the network after independent Z-score normalisation to preserve gradient stability. To comprehensively evaluate this framework, its predictive performance is benchmarked against industry-standard machine learning algorithms, including Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost). Results on variation-based unseen test data show that the hybrid model achieves competitive prediction accuracy relative to XGBoost—matching or exceeding it for Output Torque and Input Current, and remaining broadly comparable for Output Power, though XGBoost achieves substantially lower RMSE for Efficiency and Power Factor—while additionally offering transparent mathematical traceability, reducing error rates (RMSE) by 42.1% to 66.1% per parameter compared to the standard ANN. The model achieves an $R^2$ score of 0.9879 for the power factor parameter, where conventional approaches struggle. To rigorously validate the interpretability of this grey-box architecture, a global Permutation Feature Importance (PFI) analysis was conducted, showing that the GP-derived super features collectively account for 61.46% of the decision logic, outweighing the combined contribution of the raw inputs (38.54%). Furthermore, the autonomous feature engineering layer is found to reduce the learning burden on the ANN, allowing it to converge to a substantially lower error floor within the same fixed epoch budget. With an online inference time below 0.02 ms, the proposed architecture offers a robust methodological infrastructure for sustainable motor digital twins through a balanced trade-off between prediction accuracy and computational efficiency. Full article

Background/Objectives: Adolescent somatic symptoms such as headache, irritability, and sleep difficulties are common and closely tied to adolescent mental health, yet they are typically studied as a single dimension and as individual complaints detached from the psychosocial adversity surrounding the adolescent. Drawing on the nationally representative 2022 Türkiye Child Survey (n = 3523, ages 13–17), we examined whether these symptoms are patterned by the psychosocial risk and support surrounding the adolescent. Methods: Measurement models, multiple correspondence analysis, survey-weighted ordinal regression, a mixed graphical model, and machine-learning algorithms were applied in sequence. Results: Somatic symptoms were organized around a dominant general factor, alongside closely correlated physical and psychological dimensions that showed differential external associations. The three relational microsystems formed a coherent psychosocial adversity gradient along which somatic burden increased. The two dimensions were linked to different factors: peer victimization, a risk factor, and parental support, a protective factor, were associated primarily with psychological burden (odds ratios per standard deviation 1.84 and 0.88), whereas female sex and chronic illness were linked more strongly to physical burden (2.35 and 1.61); body mass index, income strain, and housing problems showed no independent associations. Regression, network, and machine-learning analyses converged on this dissociation while indicating modest individual-level predictability. Conclusions: Adolescent somatic symptoms are thus systematically patterned by psychosocial risk and support at the population level, an association that may inform monitoring and psychosocially informed assessment rather than individual prediction, and that requires longitudinal work to interpret directionally. Full article

The link between the olfactory system and psychiatric disorders is well documented; however, the neurological effects of insect-derived scents on human brain function remain unexplored. This study examined the neurological response to scented insects using functional magnetic resonance imaging (fMRI) to evaluate their potential as a novel animal-assisted intervention (AAI) for mental health. In this controlled, crossover trial, the olfactory effects of scented insects (Poecilocoris splendidulus Esaki) were compared with a control scent (alcohol) in 29 psychologically healthy adults. Brain activation differences were assessed using paired t-tests with false discovery rate correction, followed by multiple regression analyses incorporating psychometric assessments. Compared with the control scent, the insect scent elicited greater activation across sensory, emotional, and memory-related brain networks, including the thalamus, insula, hippocampus, and prefrontal cortex (p < 0.05). Furthermore, individual psychometric variations influenced these responses: higher depressive symptoms were associated with reduced activation in the anterior cingulate cortex, while elevated stress and anxiety correlated with heightened activity in the hippocampus and inferior frontal gyrus. These findings suggest that insect-derived scents induce distinct neurological responses linked to emotional processing, supporting their potential application in therapeutic interventions for mental health. Full article

Personalised three-dimensional (3D) facial avatars underpin a wide range of immersive virtual, augmented, and mixed reality (XR) experiences, yet conventional 3D capture pipelines remain prohibitively expensive and computationally demanding for prototype-stage XR applications. This study presents and evaluates a lightweight hybrid 3D Morphable Model–Convolutional Neural Network (3DMM-CNN) pipeline that reconstructs an animation-ready 3D facial mesh from a single unconstrained RGB photograph and exposes it through an interactive prototype with native export to XR-ready asset formats. A four-channel ResNet-50 backbone fuses RGB pixels with a landmark-mask channel, regresses the 3DMM shape, expression, pose, and illumination parameters, and is refined through a multi-task loss that combines 3D parameter regression, 2D landmark consistency, and image-to-mesh-to-image cycle consistency. The model is trained on a curated 2000-image subset of the LFW-People corpus and evaluated under four yaw-angle strata. The results indicate that on a held-out 400-image test set, the pipeline attains R² = 0.854, MSE = 0.022, Pearson r = 0.92, and MAPE = 10.6%, with a single-frame inference latency of 35 ms on a commodity RTX-class GPU. Robustness to head rotation improves by 29.9% at extreme poses (60–90° yaw) compared with a single-modality baseline. A Blender-integrated prototype successfully exports the reconstructed mesh as a deformation-ready asset for Unity- and Unreal-based XR engines. The proposed pipeline offers a cost-effective, real-time-capable component for XR avatar prototyping, lowering the entry barrier for small studios, immersive-learning developers, and AR/MR telepresence research. On the standard AFLW2000-3D benchmark, the pipeline additionally attains a Normalised Mean Error of 2.47% and a full-vertex reconstruction error of 1.50%, which is competitive with published lightweight baselines while retaining sub-50 ms inference latency. Full article

With the rapid advancement of artificial intelligence technologies in the era of big data, time series forecasting has become indispensable in critical fields such as environmental monitoring and financial market analysis. However, the existing forecasting models often encounter performance limitations when extracting high-dimensional features and generally cannot dynamically focus on critical information within long-term sequences. To address these challenges, this study proposes a multi-head attention fusion model (MAFM) designed to enhance the predictive accuracy and modelling capability for high-dimensional and nonlinear data across diverse application scenarios. Experiments were conducted on two heterogeneous datasets from the environmental and financial domains. After the key hyperparameters of the MAFM were optimized through an orthogonal experimental design, the model achieved coefficients of determination exceeding 0.90 on both datasets. Furthermore, the results of four comparative experiments demonstrate that the MAFM consistently outperforms traditional machine learning models, including support vector regression and extreme gradient boosting, as well as state-of-the-art deep learning models such as long short-term memory, temporal convolutional networks, and transformers. Compared with the best-performing baseline model on each sub-dataset, the MAFM reduced the mean squared error by 44.4%, 8.3%, 29.4%, and 65.5%, respectively, highlighting its superior predictive performance and strong generalization capability. In summary, the proposed MAFM provides an efficient, robust, and interpretable solution for time series forecasting tasks across multiple domains. Its outstanding performance demonstrates significant potential for practical applications in environmental monitoring, financial forecasting, and other real-world scenarios. Full article

Gastrointestinal (GI) adverse drug reactions (ADRs) are common among glucagon-like peptide-1 receptor agonist (GLP-1 RA) users and frequently contribute to treatment discontinuation and reduced therapeutic benefit. This retrospective study aimed to develop and validate an explainable artificial intelligence (XAI) model to predict GI ADR risk among GLP-1 RA users using real-world clinical data from the NIH All of Us Research Program. Adults prescribed GLP-1 RAs were identified and classified according to the occurrence of GI ADRs following treatment initiation. Multiple supervised machine learning models, including logistic regression, random forest, extreme gradient boosting (XGBoost), support vector machine, neural network, LightGBM, and CatBoost, were evaluated using demographic, socioeconomic, clinical, medication, and laboratory variables. Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1-score. A total of 8697 participants were included, of whom 59.1% experienced GI ADRs. All models demonstrated reasonable predictive performance, with AUC values ranging from 0.82 to 0.84. The XGBoost achieved discrimination of (AUC: 0.84 ± 0.01). SHapley Additive exPlanations (SHAP) identified gastroesophageal reflux disease, hemorrhoids, and elevated HbA1c as important predictors of GI ADR risk. These findings demonstrate the potential utility of explainable machine learning approaches for predicting the safety of GLP-1 RA therapy. Full article

Objectives: To evaluate bile culture microbiology and factors associated with resistance patterns in patients taken to laparoscopic cholecystectomy to better guide empiric antibiotic therapy. Methods: Prospective cohort study with a logistic regression model in a middle-income public hospital network of two institutions. Inclusion criteria were patients taken to laparoscopic cholecystectomy over 18 years of age due to benign biliary disease without other concomitant surgical procedures taken to bile culture and antibiogram testing to evaluate bile culture positivity considered as a “resistant pattern”. Results: 226 cultures tested positive for at least one microorganism, and 218 were included in the study. Overall, bile cultures classified as resistant were associated with age, comorbidities, acute signs of cholecystitis, and longer antibiotic therapy. Bile microorganisms found consisted mostly of Enterobacteriaceae. In the logistic regression model, previous ERCP, Charlson comorbidity index and duration of antibiotic therapy yielded as statistically significant (p 0.03, p 0.003 and 0.004, respectively) for presenting resistant patterns. Conclusions: Patients taken to laparoscopic cholecystectomy have a higher probability of being resistant to empirical therapy for managing acute cholecystitis if they had a higher Charlson comorbidity index, previous ERCP and longer preoperative antibiotic therapy, and thus intraoperative cultures should be considered for guided antibiotic therapy. Trial registration number: NCT06314399. Full article

Hybrid renewable energy systems (HRESs) represent a promising strategy for reducing carbon emissions across multiple sectors by integrating complementary resources such as solar, wind, hydropower, and energy storage technologies. Identifying the most suitable location for a pilot installation requires a comprehensive evaluation that balances technical performance, environmental benefits, social considerations, and economic feasibility. This study employs an enhanced multi-criteria decision analysis (MCDA) framework, supported by machine learning (ML) techniques, to assess four pilot sites developed within the HY4RES project: a rural community, an aquaculture facility, a port installation, and an agriculture network. A comprehensive set of key performance indicators (KPIs) was established to capture technical, environmental, social, and economic dimensions. These include the degree of hybridization, carbon intensity, community benefit scores, net present value, levelized cost of energy, and payback period. After collecting and normalizing the site-specific data, ML EL-SVM, decision tree, and logistic regression models as computational surrogates designed to bypass the multi-step, matrix inversion mathematical requirements of the AHP when screening massive numbers of future scenario outputs supporting consistency checks and sensitivity exploration were used, along with criterion adjustments, to refine the relative importance of each KPI. The Analytical Hierarchy Process (AHP) was employed to assess potential factors and rank the sites, with the rural site achieving the highest overall score in the system, driven by its complex four-source hybrid configuration and strong community-level benefits. The agriculture scheme ranked second, demonstrating significant potential for carbon emission reductions. The port pilot placed third, distinguished by high technical innovation but more limited social impact. The aquaculture site ranked fourth, primarily due to environmental scores, despite its economic self-sufficiency. Full article

To predict the anti-HIV-1 activity of biphenyl-DAPY-based non-nucleoside reverse transcriptase inhibitors (NNRTIs), a quantitative structure–activity relationship (QSAR) analysis was conducted. Using the heuristic method (HM) for descriptor selection, four predictive models were established: support vector regression (SVR), kernel ridge regression, linear mixed-kernel SVR, and Linear Residual Network (LRNet). Rigorous validations, including leave-one-out cross-validation, fivefold cross-validation, and Y-randomization tests, confirmed their reliability. The LRNet model exhibited the best performance, achieving an average training set $R^2$ of $0.8838\pm 0.0090$ and an average test set $R^2$ of $0.9026\pm 0.0187$ over 50 random train–test splits, with $Q_{5-fold}^2$ and $Q_{LOO}^2$ being 0.8533 and 0.8541, respectively. To further verify the robustness and generalizability of LRNet, independent validation was performed using an external dataset, where LRNet also achieved better generalization performance. The HM and LRNet models were employed to guide the design of novel compounds. Their favorable binding modes with the 1RT2 protein and pharmacokinetic properties were verified via molecular docking and in silico ADMET profiling, respectively. Furthermore, 100 ns molecular dynamics simulations demonstrated the robust dynamic stability, structural compactness, and thermodynamic convergence of the designed candidate within the 1RT2 binding pocket. This study provides a useful computational framework for the rational design and activity prediction of biphenyl-DAPY-based NNRTIs. Full article

Background: Rheumatoid arthritis (RA) is a slow progressive autoimmune disease. RA disproportionately affects women due to hormonal and immune variations. During pregnancy, hormonal and immune system changes vary drastically and may lead to RA. Traditional diagnostic techniques are blood biomarkers and clinical assessments. The above methods fail to detect RA at earlier stage due to subclinical inflammation and pregnancy-related physiological changes. Thermal imaging, as a non-invasive and radiation-free approach, can reveal temperature asymmetries across inflamed joints, offering a safer diagnostic pathway for pregnant women. Objective: In this paper, non-invasive RA detection is performed using finger, leg and hand thermal images. A pregnancy-aware rheumatoid arthritis (PARA) diagnostic framework is proposed. The PARA framework uses hybrid deep learning algorithms to classify RA inflammation states, such as normal, moderate and high. Methods: Using the PARA framework, thermal images were obtained from pregnant women. A total of 28 major bone joints were captured across four physiological states, including normal and before pregnancy. The thermal images were obtained from normal women, pregnant women, and women after pregnancy using a smartphone -based high-resolution USB thermal camera. Preprocessing was performed using bilateral, Non-Local Means (NLM), and guided filters to enhance thermal images for clarity. The guided filter preserves the edges and suppresses noise. Our proposed Denoising Convolutional Neural Network (DnCNN) algorithm was applied to preprocessed images to extract inflammation-sensitive thermal features. Finally, Multiple Linear Regression (MLR) was employed to predict the inflammation scale using the statistical values from the DnCNN-processed images. Results: The regression analysis revealed a strong correlation between thermal gradients and inflammatory severity across elbow, hand, and knee joints; i.e., the K-fold accuracy was 93.84 ± 0.71. The Modified Clinical Discord Activity Index (MCDAI) categorizes inflammation as low, moderate, and high, and these values were used in the PARA framework for inflammation level prediction supporting early clinical decision-making. Conclusion: The proposed PARA framework has high diagnostic potential to classify RA stages in pregnant women through a non-invasive and pregnancy-specific assessment. The PARA framework reduces dependency on laboratory tests and supports timely therapeutic interventions. Full article

Anomaly signals in the Attitude Determination and Control System (ADCS) of nanosatellites can significantly degrade mission performance, especially in the absence of robust Fault Detection, Isolation, and Recovery (FDIR) mechanisms. Thus, traditional threshold-based approaches, while portable and compact, may overlook subtle faults, whereby abnormal sensor signals or current spikes within the threshold may compromise the operation of the entire ADCS as a subsystem. Furthermore, the lack of interpretable detection methods further limits the development of reliable machine learning (ML) FDIR solutions. To address these limitations, this work presents a wavelet-based anomaly detection framework that introduces a two-stage hybrid architecture combining a lightweight Convolutional Neural Network (CNN) for anomaly detection with logistic regression for fault classification, both based on discrete wavelet transform (DWT) detail coefficients extracted from sensor and actuator data. The framework was validated using a statistics-based anomaly dataset for a 1U CubeSat ADCS simulated in MATLAB, in which anomalies are introduced at the component level with controlled variations in magnitude, frequency, and waveform, ensuring 99% statistical significance. Additionally, to demonstrate operational feasibility, constraints for onboard implementation were considered by executing the proposed framework in a Processor-in-the-Loop (PIL) environment. For benchmarking, lightweight detection and classification algorithms were compared, including Out-Of-Limit (OOL) and compact machine learning approaches. Finally, to identify the framework’s limitations and trace faulty events to physical phenomena, Gradient-weighted Class Activation Mapping (Grad-CAM), SHapley Additive exPlanations (SHAP), and impurity analysis were performed on the proposed algorithms as primary interpretability tools. Consequently, the results demonstrate accurate anomaly detection and identification to support both autonomous FDIR actions and ground operator decision-making. The proposed validation framework and dataset provide a reproducible basis for advancing anomaly detection onboard nanosatellites. Full article