Search Results (181)

The increasing randomness and volatility of renewable energy resources have raised higher demands for circuit breakers. Utilizing monitoring data enables more accurate condition assessment; however, the imbalance between fault and normal samples hampers the performance of machine-learning-based assessment. To address the overfitting and limited diversity of traditional oversampling methods, this paper proposes a Transformer-conditioned CWGAN-GP (TCWGAN-GP) model to generate multi-class fault samples for data augmentation. The generator of the proposed model takes random noise and class labels as input to capture the distribution characteristics of real fault samples. By combining a transformer-based generator to model inter-feature dependencies among 14 monitoring indicators and a WGAN-GP training objective with gradient penalty, the proposed approach improves training stability and synthetic sample quality. Moreover, a three-stage state assessment method based on XGBoost is developed to sequentially assess health status, fault category, and fault severity. Results demonstrate that the proposed method in the paper outperforms conventional data augmentation methods and single-stage classifiers in terms of accuracy, recall, F1-score, and online prediction efficiency. Specifically, the proposed three-stage model achieves an overall assessment accuracy of 93.10%, outperforming the single-stage XGBoost framework. In terms of online efficiency, the initial anomaly detection stage requires only 0.0041 s per sample, which is a substantial reduction compared to the 0.0241 s required by the single-stage model. Furthermore, compared to traditional Random Oversampling (ROS) and SMOTE, the TCWGAN-GP augmentation yields superior evaluation performance on fully balanced datasets, achieving a recall rate of 91.26% and an F1-score of 92.61%. Overall, the proposed TCWGAN-GP and three-stage XGBoost method contributes to addressing data imbalance challenges and improving the accuracy of circuit breaker state assessment. Full article

Against the backdrop of multi-source water transfers and increasingly frequent extreme rainfall, short-term deterioration of reservoir inflow water quality has become a key risk to intake safety, treatment operations, and urban water-supply security. Traditional assessments based on static thresholds and annual or seasonal averages often fail to identify high-risk periods at the event scale. Using continuous online monitoring data from 2021 to 2024 for the inflow of Yuqiao Reservoir, Tianjin, China, this study developed a month-specific dynamic-threshold framework and green/yellow/red risk windows and integrated a reach-wise river–reservoir routing scheme; a two-box decay model; and a three-class risk trigger into a unified analytical framework for long-term background characterization, event propagation analysis, source-contribution interpretation, and early-warning evaluation. Results show that the permanganate index (COD_Mn) exhibits an overall stable-to-declining background with pronounced wet-season pulses, whereas total nitrogen (TN) and total phosphorus (TP) remain at moderate-to-high levels, with yellow/red risk windows clustering markedly in the wet season. In typical red and yellow events, nitrogen contributions from upstream control sections progressively accumulate toward the reservoir inlet along the river–reservoir cascade system, whereas in some events the residual contribution from unmonitored near-inlet inflows becomes dominant. The COD_Mn-based three-class trigger achieves an overall accuracy of approximately 71.5% and shows comparatively strong identification of yellow-level risk, while remaining conservative for red-level alarms. These findings indicate that coupling month-specific dynamic thresholds with event-scale routing-decay analysis and trigger-based classification can support inflow monitoring, intake-risk early warning, and coordinated operation of key upstream reaches and near-reservoir control zones in water-transfer–reservoir integrated systems. Full article

Leakage identification in district heating networks is challenging because leakage-induced transient pressure waves often overlap with pressure disturbances triggered by routine operations such as valve regulation, pump speed variation, and emergency shut-off. In addition, the scarcity of high-quality labeled leakage samples limits the robustness of data-driven models under small-sample conditions. To address these issues, this study proposes a compact and moderately interpretable framework for multi-condition identification from transient pressure-wave signals, integrating signal preprocessing, handcrafted statistical feature extraction, multiclass ReliefF-based feature selection, and class-wise generative adversarial network augmentation in the selected feature space. A dataset containing four representative conditions, namely leakage, valve regulation, pump speed regulation, and emergency valve shut-off, was constructed using an integrated indoor district heating network testbed. After Hampel-based spike suppression and zero-phase Butterworth band-pass filtering within 0.5 to 300 Hz, time- and frequency-domain statistical features were extracted, and a compact subset was selected by multiclass ReliefF. A class-wise generative adversarial network was then used to augment the training set in feature space, while all evaluations were performed strictly on real samples. The results show that feature-space augmentation improves robustness and generalization under operational disturbances and noise. Using random forest as the representative classifier, Accuracy and Macro-F1 increased from 0.960 to 0.985, while leakage recall improved from 0.920 to 0.980. Further comparisons confirmed that the ReliefF-selected subset outperformed representative alternatives such as LASSO and mRMR. Overall, the proposed framework provides an effective solution for distinguishing leakage events from operational disturbances and offers practical support for online monitoring and intelligent operation of district heating networks. Full article

Monitoring students’ on-screen availability is increasingly critical for analyzing participation patterns in synchronous online learning, especially under videoconferencing conditions characterized by compressed video streams, low-resolution face regions, fluctuating bandwidth, and dynamically reconfigured grid layouts. This study introduces a practical computer vision pipeline that integrates deep learning-based face detection, lightweight embedding-based identity matching, and frame-level temporal aggregation to estimate students’ visual presence (VP) during live online classes. A real-world dataset comprising 27 participants and 16,200 frames was collected under authentic conditions, including codec compression, variable image quality, and dynamic layout changes. Four widely used face detection models (Haar Cascade, DSFD, MTCNN, and YuNet) were benchmarked on noisy and low-quality images. Quantitative evaluation on a manually annotated subset of 270 frames demonstrates that MTCNN and YuNet yield lower average VP estimation errors (27.63% and 22.20%, respectively) compared to Haar Cascade (75.34%) and DSFD (47.14%), with YuNet also achieving the shortest average processing time of 0.069 s per frame. While the pipeline is intentionally streamlined to facilitate practical use by instructors, the study provides clearly defined steps and parameter settings, establishing a reproducible procedure for benchmarking face detection performance in synchronous online class environments. Full article

To meet the energy-saving requirements of user equipment (UE) operating in Radio Resource Control idle/inactive states (RRC_IDLE/RRC_INACTIVE) in the 3rd-Generation Partnership Project (3GPP) 5G-Advanced (5G-A) networks, the New Radio (NR) downlink paging procedure relies on periodic monitoring and frequent synchronization signal block (SSB) measurements, which wastes energy when no paging arrivals occur. Meanwhile, heterogeneous Quality of Service (QoS) constraints make it difficult for fixed-parameter Idle Discontinuous Reception and Paging Early Indication mechanisms (IDRX/PEI) to balance energy, delay, and reliability. This paper develops a UAV-assisted 5G-A paging control framework that maps services into multiple QoS classes and models QoS violation risk and system energy consumption under unified accounting, including UE monitoring/reception energy and unmanned aerial vehicle (UAV) forwarding energy. We then propose a QoS-aware risk-driven paging strategy: an offline Long Short-Term Memory (LSTM) predictor is trained to estimate the time-to-next-arrival (TTNA) of paging events and produce a bounded urgency/risk signal to initialize class-dependent thresholds, while online triggering and QoS-feedback-based threshold adaptation regulate the empirical violation rate toward target constraints under varying loads, enabling a controllable energy–delay trade-off. A simulation-based evaluation is conducted to compare the proposed method with representative baselines (Enhanced Paging Monitoring (EPM), Split Paging Occasion (SPOP), and Predicted Paging Early Indication (PPEI)) and to examine the impact of SSB overhead and UAV relaying on the energy–delay–reliability trade-offs. Full article

Supply Chain Management (SCM) has received considerable attention from the industrial community in recent decades. SCM continues to be an interesting and relevant research topic in many business areas such as revealing supply chain integration benefits, uncertainty and risk mitigation methods, decision-making and optimization methodologies, etc. In current supply chain management, huge volumes of data are being developed each second, and emerging technologies such as Radio Frequency Identification (RFID) have amplified the availability of online data. Using Artificial Intelligence (AI) methods that go beyond simply using the huge volume of online data enables Supply Chain (SC) managers to monitor everything in a timely fashion. There are several aspects of an SC that AI—and specifically Artificial Neural Networks (ANNs)—can be applied to better help them manage and optimize. This study aims to review state-of-the-art ANNs and Deep Neural Networks (DNNs) in the field of supply chain management. One hundred high-quality research studies that applied ANNs in supply chain management are reviewed and categorized into four classes: performance optimization, supplier selection, forecasting, and inventory management studies. Our study shows that there is a significant possibility that we could use ANNs and DNNs to better manage supply chains. Across the reviewed studies, neural networks are frequently reported to improve predictive performance and support monitoring/control in complex, nonlinear supply chain settings, often complementing traditional operations research approaches. Finally, the limitations of ANN models and the possibilities for future studies are presented at the end of this study. Full article

Urban airborne laser scanning (ALS) point clouds cover extensive geographical areas, rendering dense point-level annotation economically prohibitive and limiting the feasibility of fully supervised learning. In weakly supervised settings for urban ALS data, the natural long-tailed class distribution—where ground and building points dominate and smaller objects are rare—combined with the use of fixed pseudo-label thresholds under sparse annotations exacerbates confirmation bias and increases prediction uncertainty. This ultimately restricts the effective utilization of unlabeled data during training. To overcome these challenges, we propose a pseudo-label confidence-calibrated curriculum learning framework designed for weakly supervised ALS point cloud classification. The framework introduces a confidence-aware self-adaptive soft gating (CSS) mechanism that dynamically adjusts category-specific thresholds online using exponential moving average statistics and scene-aware normalization, eliminating the need for manual scheduling while improving pseudo-label quality. In addition, a reliability-driven soft selection (RSS) constraint is incorporated, in which each point is assigned a comprehensive reliability score that integrates prediction confidence, entropy clarity, and cross-augmentation consistency, enabling adaptive soft weighting to replace hard pseudo-label selection and achieve more balanced sample utilization. These components are further integrated into a unified pseudo-label confidence-calibrated curriculum learning framework (P3CL) that progressively shifts the model’s focus from high-certainty samples to more ambiguous ones, effectively mitigating confirmation bias. Extensive experiments on three public ALS benchmarks demonstrate that the proposed method consistently outperforms existing weakly supervised approaches and achieves competitive performance compared with several fully supervised models. Full article

Background and Objectives: Abortion on request, contraceptive access barriers, and mental health may jointly shape women’s quality of life (QoL). We examined how abortion history, structural barriers, and psychosocial factors relate to modern contraceptive use, depressive and anxiety symptoms, and QoL among women attending a Romanian tertiary center. Methods: We conducted a single-center observational study combining retrospective chart review with an online survey of 200 women aged 18–45 years. Validated instruments (Patient Health Questionnaire-9 [PHQ-9], Generalized Anxiety Disorder-7 [GAD-7], World Health Organization Five-Item Well-Being Index [WHO-5], and World Health Organization Quality of Life–BREF [WHOQOL-BREF]) and indices of access barriers, perceived stigma, and social support were used. Analyses included multivariable regression, structural equation modelling, latent class analysis, and moderation analysis. Results: Overall, 55.0% of women reported ≥1 abortion on request. Compared with those without abortion history, they were older (31.2 ± 4.9 vs. 26.8 ± 4.8 years, p < 0.001), more often had lower levels of education (51.8% vs. 33.3%, p = 0.013), and were less likely to use modern contraception at last intercourse (52.7% vs. 71.1%, p = 0.012). PHQ-9 (8.8 ± 4.0 vs. 7.3 ± 4.3) and GAD-7 (7.0 ± 3.2 vs. 5.7 ± 3.4) scores were higher (both p = 0.010), while QoL was lower (55.4 ± 8.1 vs. 59.5 ± 7.8, p < 0.001). In adjusted models, access barriers (OR per point = 1.3, 95% CI 1.1–1.6), but not abortion history, predicted non-use of modern contraception. QoL correlated strongly with PHQ-9 (r = −0.6) and WHO-5 (r = 0.5; both p < 0.001). Latent class analysis identified a “high-barrier, distressed, abortion-experienced” profile with the poorest mental health and QoL. Conclusions: Structural access barriers and current depressive and anxiety symptoms, rather than abortion history alone, were key correlates of contraceptive gaps and reduced QoL, underscoring the need for integrated reproductive and mental health care. Full article

This paper introduces Saudi Dialects Cyber Violence Detection (SD-CVD) corpus, a large-scale, class-balanced Saudi-dialect corpus for fine-grained cyber violence detection on online platforms. The dataset contains 88,687 Saudi Arabic tweets annotated using a three-level hierarchical scheme that assigns each tweet to one of 11 mutually exclusive classes, covering benign sentiment (positive, neutral, negative), cyberbullying, and seven hate-speech subtypes (incitement to violence, gender, national, social class, tribal, religious, and regional discrimination). To mitigate the class imbalance common in Arabic cyber violence datasets, data augmentation was applied to achieve a near-uniform class distribution. Annotation quality was ensured through multi-stage review, yielding excellent inter-annotator agreement (Fleiss’ κ > 0.89). We evaluate three modeling paradigms: traditional machine learning with TF–IDF and n-gram features (SVM, logistic regression, random forest), deep learning models trained on fixed sentence embeddings (LSTM, RNN, MLP, CNN), and fine-tuned transformer models (AraBERTv02-Twitter, CAMeLBERT-MSA). Experimental results show that transformers perform best, with AraBERTv02-Twitter achieving the highest weighted F1-score (0.882) followed by CAMeLBERT-MSA (0.869). Among non-transformer baselines, SVM is most competitive (0.853), while CNN performs worst (0.561). Overall, SD-CVD provides a high-quality benchmark and strong baselines to support future research on robust and interpretable Arabic cyber-violence detection. Full article

Online chess platforms generate vast amounts of game data, presenting opportunities to analyze match outcomes using machine learning approaches. This study develops and compares four machine learning models to classify chess game results (White win, Black win, or Draw) by integrating player rating information with game dynamic metadata. We analyzed 11,510 complete games from the Lichess platform after preprocessing a dataset of 20,058 initial records. Seven key features were engineered to capture both pre-game skill parameters (player ratings, rating difference) and game complexity metrics (game duration, turn count). Four machine learning algorithms were implemented and optimized through grid search cross-validation: Multinomial Logistic Regression, Random Forest, K-Nearest Neighbors, and Histogram Gradient Boosting. The Gradient Boosting classifier achieved the highest performance with 83.19% accuracy on hold-out data and consistent 5-fold cross-validation scores (83.08% ± 0.009%). Feature importance analysis revealed that game complexity (number of turns) was the strongest correlate of the outcome across all models, followed by the rating difference between opponents. Draws represented only 5.11% of outcomes, creating class imbalance challenges that affected classification performance for this outcome category. The results demonstrate that ensemble methods, particularly gradient boosting, can effectively capture non-linear interactions between player skill and game length to classify chess outcomes. These findings have practical applications for chess platforms in automated content curation, post-game quality assessment, and engagement enhancement strategies. The study establishes a foundation for robust outcome analysis systems in online chess environments. Full article

The growth of electronic word-of-mouth (eWOM) on digital platforms has heightened the need to distinguish authentic user-generated content from covert promotional material. This study proposes an integrated framework combining Natural Language Processing (NLP), machine learning, and Latent Dirichlet Allocation (LDA) to classify sentiment and detect advertising features in online game reviews. Reviews from the Steam platform were analyzed using Support Vector Machine (SVM), Decision Tree, and Naïve Bayes classifiers, with class imbalance addressed through SMOTE and SMOTE–Tomek techniques. The SMOTE-augmented SVM achieved the highest performance, with 98.18% overall accuracy and 97.52% negative sentiment detection. LDA and Quality Function Deployment (QFD) further uncovered latent promotional themes, providing insights into how advertising elements manifest in positive reviews and how negative feedback reflects genuine user concerns. The framework assists platform managers in enhancing eWOM credibility and supports marketers in designing data-driven advertising strategies. By bridging sentiment analysis with covert marketing detection, this research contributes a novel methodological approach for assessing review trustworthiness, improving transparency, and fostering consumer trust in digital information environments. Full article

Background: Population ageing is a global challenge, prompting ageing-in-place policies in Hong Kong to support community-dwelling older adults while reducing healthcare costs. Yet, their impact on health-related quality of life (HRQoL) remains underexplored amid Hong Kong’s long life expectancy and growing older population. Traditional surveys are costly and time-consuming, while routinely collected registration data offers a large, efficient source for health insights. This study uses enhanced administrative data to track HRQoL trajectories and inform policy. Methods: This is a prospective, open-ended longitudinal study, enrolling adults aged 50 or older from a collaborating non-governmental organization in Hong Kong’s Southern District. Data collection, started in February 2021, occurs annually via phone and face-to-face interviews by trained social workers and volunteers using a standardized questionnaire to assess individual (e.g., socio-demographics), environmental (e.g., social support via Lubben Social Network Scale-6), biological (e.g., chronic illnesses), functional (e.g., cognition via Montreal Cognitive Assessment), and HRQoL (e.g., EQ-5D-5L) factors. A secure online system links health and service use data (e.g., service utilization like community care visits). Analysis employs descriptive statistics, group comparisons, correlations, growth modelling to identify health trajectories, and structural equation modelling to test a revised quality-of-life framework. Sample size (projected 470–580 after two follow-ups from a 2321 baseline) is based on power calculations: 300–500 for latent class growth analysis (LCGA) class detection and 200–400 for structural equation modelling (SEM) fit (e.g., RMSEA < 0.06) at 80% power/α = 0.05, simulated via Monte Carlo with a 50–55% attrition. Discussion: This is the first longitudinal HRQoL study in Hong Kong using enhanced non-governmental organization (NGO) administrative data, integrating social–ecological and HRQoL models to predict trajectories (e.g., stable vs. declining mobility) and project care demands (e.g., increase in in-home care for frailty). Unlike prior cross-sectional or inpatient studies, it offers a scalable model for NGOs, informing ageing-in-place policy effectiveness and equitable geriatric care. Full article

The Vehicular Internet of Things (V-IoT) has emerged as a cornerstone of next-generation intelligent transportation systems (ITSs), enabling applications ranging from safety-critical collision avoidance and cooperative awareness to infotainment and fleet management. These heterogeneous services impose stringent quality-of-service (QoS) demands for latency, reliability, and fairness while competing for limited and dynamically varying spectrum resources. Conventional schedulers, such as round-robin or static priority queues, lack adaptability, whereas deep reinforcement learning (DRL) solutions, though powerful, remain computationally intensive and unsuitable for real-time roadside unit (RSU) deployment. This paper proposes a lightweight and interpretable reinforcement learning (RL)-based spectrum management framework for Vehicular Internet of Things (V-IoT) networks. Two enhanced Q-Learning variants are introduced: a Value-Prioritized Action Double Q-Learning with Constraints (VPADQ-C) algorithm that enforces reliability and blocking constraints through a Constrained Markov Decision Process (CMDP) with online primal–dual optimization, and a contextual Q-Learning with Upper Confidence Bound (Q-UCB) method that integrates uncertainty-aware exploration and a Success-Rate Prior (SRP) to accelerate convergence. A Risk-Aware Heuristic baseline is also designed as a transparent, low-complexity benchmark to illustrate the interpretability–performance trade-off between rule-based and learning-driven approaches. A comprehensive simulation framework incorporating heterogeneous traffic classes, physical-layer fading, and energy-consumption dynamics is developed to evaluate throughput, delay, blocking probability, fairness, and energy efficiency. The results demonstrate that the proposed methods consistently outperform conventional Q-Learning and Double Q-Learning methods. VPADQ-C achieves the highest energy efficiency (≈$8.425\times {10}^7$ bits/J) and reduces interruption probability by over $60\%$, while Q-UCB achieves the fastest convergence (within ≈190 episodes), lowest blocking probability (≈0.0135), and lowest mean delay (≈0.351 ms). Both schemes maintain fairness near $0.364$, preserve throughput around 28 Mbps, and exhibit sublinear training-time scaling with $\mathcal{O}\left(1\right)$ per-update complexity and $\mathcal{O}\left(N^2\right)$ overall runtime growth. Scalability analysis confirms that the proposed frameworks sustain URLLC-grade latency (<0.2 ms) and reliability under dense vehicular loads, validating their suitability for real-time, large-scale V-IoT deployments. Full article

Extracorporeal membrane oxygenation (ECMO) is increasingly used to support critically ill adults with severe cardiac or respiratory failure, but ECMO circuits and the physiological disturbances of critical illness significantly alter drug pharmacokinetics (PK) and pharmacodynamics (PD), complicating dosing and monitoring. This narrative review synthesizes current clinical evidence on ECMO-related PK/PD alterations and provides practical guidance for optimizing pharmacotherapy in adult intensive care. A structured literature search (January–May 2025) was conducted across PubMed/MEDLINE, EMBASE, Scopus, Cochrane Library, Sage Journals, ScienceDirect, Taylor & Francis Online, SpringerLink, and specialized databases, focusing on seven therapeutic classes commonly used in ECMO patients. Eligible studies included clinical trials, observational studies, systematic reviews, and practice guidelines in adults, while pediatric and preclinical data were excluded. Evidence quality varied substantially across drug classes. Hydrophilic, low-protein-bound agents such as β-lactams, aminoglycosides, fluconazole, and caspofungin generally showed minimal ECMO-specific PK alterations, with dose adjustment mainly driven by renal function. Conversely, lipophilic and highly protein-bound drugs including fentanyl, midazolam, propofol, voriconazole, and liposomal amphotericin B exhibited substantial circuit adsorption and variability, often requiring higher loading doses, prolonged infusions, and rigorous therapeutic drug monitoring. No ECMO-specific data were identified for certain neuromuscular blockers, antivirals, and electrolytes. Overall, individualized dosing guided by therapeutic drug monitoring (TDM), organ function, and validated PK principles remains essential to optimize therapy in this complex population. Full article

Student disengagement remains a major barrier to effective learning in synchronous online classrooms, where lack of interaction, limited feedback, and screen fatigue often contribute to passive participation. Despite the growing use of generative AI, there is a notable lack of empirical research investigating the application of generative AI in addressing engagement challenges in synchronous online sessions. This study introduces VClass Engager, a novel experimental system that utilizes generative AI to foster student participation and deliver instant, personalized feedback during live virtual sessions. The system integrates several features, including instant analysis of student answers to chat questions, real-time AI-generated feedback, and a leaderboard that displays students’ cumulative scores to promote sustained engagement. To assess its effectiveness, the system was evaluated across multiple courses. Engagement was measured by tracking participation in three in-class formative questions, and response quality was analyzed using a cumulative link mixed model (CLMM). In addition, a post-session survey captured students’ perceptions regarding usability, motivational impact, and feedback quality. Results demonstrated statistically significant increases in student participation and response quality in sessions using VClass Engager compared to baseline sessions. Survey responses revealed high levels of satisfaction with the system’s ease of use and motivational aspects. By combining AI and gamification, this study provides early empirical evidence for a promising approach to enhancing engagement in synchronous online learning. Full article