Search Results (3,189)

The proliferation of unsolicited short messages (SMS spam) poses persistent challenges to mobile communication security and user privacy. This study presents a systematic benchmarking and analytical investigation of classical machine learning approaches for SMS spam detection, focusing on the impact of text feature representation under imbalanced short-text conditions.In practical SMS filtering systems, minimizing false positives (i.e., incorrectly blocking legitimate messages) is a critical operational constraint. Therefore, beyond overall accuracy, precision and specificity are emphasized to ensure reliable preservation of legitimate communication. Using the SMSSpamCollection dataset (5574 messages: 747 spam and 4827 ham), seven feature representation techniques were evaluated in combination with six widely adopted classifiers, resulting in 42 configurations assessed under 10-fold cross-validation. The results demonstrate that feature representation plays a more critical role than classifier complexity. Character-level 3-grams combined with Logistic Regression achieved the best overall performance, reaching 98.55% accuracy, with 98.55% precision and 90.50% recall for the spam class (F1-score = 94.32%), and 0.9893 AUC. Linear SVM produced comparable results, highlighting the effectiveness of linear models when paired with expressive representations. Beyond reporting performance metrics, this study analyzes feature–classifier interaction patterns and clarifies practical trade-offs between precision, recall, and computational efficiency. The findings provide reproducible baselines and structured guidance for designing efficient SMS spam filtering systems. Full article

Tracking structural behavior is critically important to reduce maintenance and repair costs. Structural Health Monitoring (SHM) aims to evaluate the structural integrity, detect damage or abnormalities, and estimate overall safety. The integration of Machine Learning techniques has significantly advanced SHM by enabling the identification of deterioration patterns through sensor data analysis. This study focuses on classifying different vibration patterns recorded under various excitation scenarios (ambient, transient, and forced) using sensors installed directly on a 3-DoF structure. The proposed approach used a two-dimensional convolutional neural network (2D-CNN) trained on vibration image patterns generated from vibration signal scalogram images. To address dataset imbalance, stratified 5 × 3 Nested cross-validation and multiple performance metrics were computed to ensure robust evaluation. The proposed method was compared with single-sensor scalogram approaches and baseline models, including Support Vector Machines (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), One-Dimensional Convolutional Neural Network (1D-CNN), and Long Short-Term Memory (LSTM) models, incorporating class-weighting strategies. Additionally, the contribution of the Total Energy Delivered by Sensor (TES) feature was evaluated for SVM, RF, and XGBoost models. The 2D-CNN model achieved superior performance in identifying excitation types associated with structural dynamic behavior, highlighting its effectiveness for structural vibration pattern recognition in SHM applications. Full article

Temperate forests play a key role in biodiversity conservation, climate regulation, and the provision of ecosystem services. However, land-use changes and urban expansion have intensified landscape fragmentation processes, reducing ecological connectivity and ecosystem functionality. Despite the importance of community-owned forests in northern Mexico, evaluations of landscape configuration within these territories remain limited. This study compared land-use and land-cover patterns and fragmentation metrics in four community-managed ejidos in Durango, Mexico, using Landsat imagery from 2000 and 2020. Land-cover maps were produced through supervised classification with a Random Forest algorithm and validated using standard accuracy metrics. Landscape composition, configuration and connectivity were assessed at class and landscape levels using a set of spatial metrics calculated with FRAGSTATS. The results reveal contrasts among ejidos. Ciénega de los Caballos and Navajas show greater representation of secondary vegetation accompanied by changes in patches and edge densities. San retains a more cohesive configuration with comparatively higher aggregation and connectivity, whereas El Tunal y Anexos exhibit stronger subdivision and lower connectivity. These outcomes emphasize the value of spatial metrics for identifying differences in landscape structure between observation years and for supporting comparative assessment in community-managed forest territories. The study provides spatially explicit information that may assist territorial planning and forest management at this scale. Full article

Background: Flavonoids and phenolic acids are recognized for their diverse therapeutic potential, yet their translation into clinical applications remains limited by varying bioavailability and fragmented characterization across databases. A systematic integrative approach is needed to comprehensively evaluate these compounds’ drug-likeness properties based on computational metrics, molecular interactions, and dietary sources within a unified framework. Methods: We analyzed 954 compounds (715 flavonoids, 239 phenolic acids) by integrating data from PhytoHub, Phenol-Explorer, ChEMBL, and FoodDB databases. Drug-likeness was assessed using established metrics, including QED (Quantitative Estimate of Drug-likeness) and DataWarrior drug-likeness scores. Molecular interaction patterns were characterized through ChEMBL activity data, and food source distributions were systematically mapped across major food groups. Results: Drug-likeness assessment revealed complementary evaluation patterns between QED (mean = 0.48 ± 0.24) and DataWarrior scores (mean = −2.46 ± 4.38), with moderate inter-correlation (r = 0.41), indicating that each metric captures distinct aspects of molecular properties. Isoflavones demonstrated the most favorable drug-likeness profiles (mean QED: 0.62 ± 0.18). Molecular interaction analysis demonstrated significantly higher binding affinities for flavonoids (mean ChEMBL activity score: 7.26 ± 1.09) compared to phenolic acids (6.98 ± 0.94, p = 0.014), with flavonoids targeting a broader range of proteins (67 unique targets vs. 33 for phenolic acids). Food source mapping identified herbs and spices as the richest sources (up to 14,500 mg/kg), followed by fruits (40,490 mg/kg total) and teas (37,101 mg/kg total), with distinct compound distribution patterns across food groups. Conclusions: This integrative cross-database approach provides a comprehensive characterization framework for flavonoids and phenolic acids, combining established drug-likeness metrics, molecular interaction analysis, and dietary source mapping. The methodology establishes a systematic foundation for compound evaluation in drug development and nutritional research. Full article

Graph-based clustering is a fundamental task in unsupervised machine learning and has been extensively applied to complex data mining scenarios, such as pattern recognition and data classification. However, most existing graph clustering algorithms still face significant challenges, including low graph learning efficiency, poor adaptability to datasets with large numbers of samples and clusters, and inevitable accuracy loss caused by post-processing steps. To effectively tackle these critical challenges and enhance clustering performance, we propose a novel Fast Discrete Clustering algorithm integrated with Local Graph Learning, namely FDC-LGL. Based on the classical normalized cut criterion, the proposed algorithm innovatively integrates a Local Graph Learning module into the clustering objective function, efficiently and reliably learning graph structures by introducing second-order neighbor constraints. It directly outputs accurate clustering results through a discrete indicator matrix, thereby eliminating the need for additional post-processing. Extensive comparative experiments conducted on synthetic datasets, medium-scale real-world datasets, and large-scale real-world datasets demonstrate that FDC-LGL is significantly superior to other state-of-the-art graph clustering algorithms in terms of key evaluation metrics, including clustering accuracy (ACC), normalized mutual information (NMI), and the adjusted rand index (ARI), as well as computational efficiency. Full article

Micronutrient deficiencies remain a persistent challenge to global health and food security, particularly in low- and middle-income countries where evidence-based strategies are urgently needed. Biofortification of staple crops has been promoted as a complementary intervention to supplementation and food fortification, but its effective implementation requires locally relevant studies. Such evidence is essential because the performance and adoption of biofortified crops depend on context-specific factors, including crop varieties, soil micronutrient dynamics, dietary patterns, cultural acceptability, and bioavailability, which limit the transferability of findings across settings. This perspective examines whether countries with the highest micronutrient burdens generate sufficient local research to inform biofortification policy decisions. We conducted a bibliometric mapping of peer-reviewed literature indexed in Scopus and compared country-level publication counts with indicators of iodized salt coverage, zinc deficiency, and childhood anemia, which were selected because they are prioritized metrics in global health and food security. From 776 eligible articles, most publications originated from a small group of high- and middle-income countries, whereas regions facing the greatest nutritional burdens, including parts of Sub-Saharan Africa and South Asia, contributed little to the scientific output. Countries with low iodized-salt coverage, high zinc deficiency, or childhood anemia above 40% frequently showed zero or minimal publications. This misalignment suggests that countries facing the greatest nutritional vulnerabilities may be underrepresented in the indexed scientific literature. These findings highlight the value of further strengthening research participation and visibility in high-burden settings to ensure that the evidence base more accurately reflects global needs. Full article

Mesoscale eddies are highly active around the Kuroshio in the East China Sea (ECS), serving as a crucial component of the ECS’s complex dynamic environment. However, the long-term variation in mesoscale eddies in this region has not been systematically investigated. Based on daily satellite altimeter data spanning from January 1993 to December 2023, this study comprehensively investigates the trend characteristics of mesoscale eddies in the ECS during this period, using eddy metrics such as Eddy Kinetic Energy (EKE) and eddy polarity probability. EKE in the ECS is primarily high around the Kuroshio, exhibiting a significant increasing trend. This upward trend is more pronounced in summer, autumn, and winter, all of which pass the significance test. From the statistics of coherent mesoscale eddies, cyclonic and anticyclonic eddies show opposite trend characteristics: cyclonic eddies display trends of decreasing number and weakening intensity, while anticyclonic eddies exhibit trends of increasing number and strengthening intensity. Energy transfer from the background flow makes a certain contribution to the aforementioned trends, but is relatively complex. The opposing trend characteristics exhibited by eddies of different polarities are related to the work done by the upper-ocean wind field. The nonuniform responses of wind-related changes in cyclonic and anticyclonic eddies could affect the regional patterns of ocean circulation and biogeochemical responses to future climate change. Full article

Despite the improvements in tool development for DNA methylation analysis, there is a lack of a consensus on computational and statistical models used for differentially methylated cytosine (DMC) identification. This variability complicates the interpretation of findings and raises concerns about the reproducibility and biological significance of the detected results. In this regard, here we conducted a comparative evaluation of edgeR and methylKit tools to assess their performance, concordance, and biological relevance in detecting DMCs following a morphine exposure model in mouse embryonic stem cells (mESCs). Both pipelines were applied to the same WGBS dataset (GEO accession number: GSE292082), and concordance was calculated at both single-base and gene levels. Although the total number of DMCs identified differed between tools, both pipelines detected a global hypomethylation pattern. Genomic distribution analysis revealed that DMCs predominantly localized to intergenic and intronic regions, as well as to open sea regions. Despite differences in sensitivity, both pipelines demonstrated moderate concordance at the DMC level (~56%) and high concordance at the gene level (~90%), identifying largely overlapping sets of differentially methylated genes (DMGs). Comparative assessments further showed that the choice of statistical metric can influence the perceived magnitude of biological effects. Sensitivity analyses indicated that threshold selection and normalization methods influence DMC detection, whereas aggregation at gene level reduces discrepancies. Overall, our findings underscore the complementary strengths of methylKit and edgeR and highlight the importance of careful tool selection for epigenetic studies. As a conclusion, we recommend integrating both pipelines to ensure a balanced interpretation of effect sizes, particularly in studies with complex experimental designs. Full article

Lithographic patterning continues to evolve under the dual pressure of ever-finer features and manufacturable, cost-effective integration. Beyond headline resolution, industrial adoption is increasingly determined by a small set of coupled metrics: throughput, overlay (registration), defectivity, and cost, as well as by how these trade-offs shift with materials, substrate form factors, and integration flows. Here, we review lithographic techniques across three eras: traditional methods (pre-1990s), non-conventional innovations (1990s), and contemporary advancements (post-2000s), with an explicit goal that goes beyond compilation. Specifically, we provide a decision framework for interpreting each method using the same manufacturing-relevant criteria. For each class of technique, we summarize the operating principle and representative process routes, then map the dominant bottlenecks to the metric that ultimately limits scale-up. This cross-cutting lens clarifies why many emerging methods are compelling at the physics level yet remain constrained at the system level, where process windows, in-line control, and compatibility with existing fabrication ecosystems govern viability. By connecting mechanism-level innovation to manufacturing-level constraints, this review offers practical guidance for researchers and engineers seeking to position nanolithography options for applications ranging from high-volume semiconductor production to agile prototyping and materials- or substrate-limited devices. Full article

This study combined experimental analysis with deep learning to investigate the effects of curing age, steel slag content, and gradation composition on the mechanical properties of cement-stabilized steel slag (CSSS). The strength evolution patterns and underlying microscopic mechanisms were systematically elucidated. Experimental results showed that CSSS strength grows nonlinearly with curing age, with optimal mechanical performance achieved at a 60% steel slag content. The microstructural evolution characterized by SEM-EDS and XRD revealed that steel slag incorporation promotes the formation of AFt and densifies the gel network. In later curing stages, natural carbonation of Ca(OH)₂ and secondary hydration of reactive steel slag components produce CaCO₃ and additional C-S-H gel, which fill pores and significantly enhance long-term strength. A CNN-GRU-Attention model was developed to predict the unconfined compressive strength (UCS) and splitting tensile strength (STS) of CSSS. In a single data split, the model achieved R² values of 0.9875 for UCS and 0.9911 for STS, with RMSEs of 0.2577 MPa and 0.0234 MPa, and MAEs of 0.2059 MPa and 0.0184 MPa, outperforming all benchmark models. Under rigorous 5 × 5 repeated cross-validation, it maintained the highest average R² (UCS: 0.9417, STS: 0.9329) and the lowest error metrics, confirming its robustness and generalization capability. SHAP and Pearson correlation analyses identified cement content as the primary strength determinant, while steel slag content exhibited a threshold effect, highlighting the importance of prudent gradation control in practical engineering. This study provides both a theoretical foundation and a methodological framework for analyzing variable interactions and predicting the strength development of CSSS. Full article

This study examines the vocal directivity of singing in Greek across three stylistically diverse genres—operatic/classical, modern/pop, and Byzantine chant—performed under realistic, unconstrained conditions. Directivity data was captured in a hemi-anechoic environment using a 29-microphone hemispherical array, in a setup that allowed singers to make natural, performance-related micro-movements. The applied analysis framework combined sound projection (magnitude of radiated energy across space and frequency) and radiation patterns (normalized spatial distribution) with three established directivity metrics: Horizontal Directivity Index, Front-to-Back Ratio, and Upward-to-Downward Ratio. Results show that while directional shape remains largely consistent across styles and sexes, projection intensity varies systematically as a function of both. Male pop singers exhibit the strongest low-frequency output (125–500 Hz), while female classical and male pop/Byzantine singers display greater frontal focus in the 1–2 kHz range. Classical singers tend toward more balanced projection profiles. Beyond the release of publicly available datasets—including the first directivity measurements of Byzantine chant—this study introduces a structured analysis framework and offers comparative findings that inform vocal science, pedagogy, and spatial audio applications. Full article

Unmanned aircraft systems for regional operations are assumed to frequently operate at non-towered airports, where routine integration remains challenging due to limited separation principles and partially observable manned traffic intent. This research investigates tactical procedures for integrating unmanned aircraft into non-towered airport environments, where unmanned aircraft must interact with manned traffic under procedural constraints. A simulation framework is developed that combines historical traffic data with standard traffic pattern procedures and rule-based decision-making to integrate unmanned aircraft at non-towered airports. The simulation logic includes detection of manned traffic activities, rule-based queuing, and airspace capacity constraints. By varying detection look-ahead times (60/120/180 s) and unmanned aircraft traffic rates (15/30 min), the simulation quantifies terminal airspace accessibility and derives metrics that capture throughput (no conflict versus deconflicted holding flights), delay propagation (holding minutes and holding orbit counts), concept feasibility (aborted/denied holdings), and altitude band utilization. The results show a consistent safety versus throughput trade-off with longer look-ahead times increasing holding demand but reducing the share of aborted holdings, while higher traffic volumes amplify holdings and delay. Holdings are predominantly conducted in the lowest available holding altitude at 2500 feet above the ground, with occasional multi-layer use to handle traffic peaks. Full article

The widespread adoption of Learning Management Systems (LMSs) in higher education has not necessarily led to profound pedagogical transformations, raising questions about how digital technologies are actually integrated into teaching practices. This study aims to describe and map the techno-pedagogical configurations used by instructors at the University of Extremadura (Spain) to identify teaching patterns and characterize course design within a virtual campus. Using log data from 12,361 Moodle courses during the 2021–2022 academic year, a K-means clustering analysis was applied to classify courses based on their use of digital tools and resources. The analysis identified five distinct clusters ranging from traditional models with minimal LMS use (67.37%) to advanced innovation configurations (3.98%), including audiovisual-based innovation, participative traditional models, and repository-focused approaches. The results indicate that the implementation of techno-pedagogical strategies is highly heterogeneous and that courses classified as more innovative do not consistently produce better academic outcomes. These findings suggest that for most faculty, the LMS functions as digital support for traditional face-to-face teaching rather than as a driver of pedagogical transformation, highlighting the need for teacher-centered analytics capable of capturing instructional design patterns beyond mere behavioral metrics. Full article

Interchange merging areas are critical safety hotspots in urban road networks, where complex vehicle interactions challenge traffic safety and efficiency. Improving safety performance at these locations is essential for developing sustainable, resilient, and intelligent urban transportation systems. To overcome the limitations of single-driver simulators, this study developed a multi-driver simulation platform based on Unity3D (Version 2022.3.1f1c1), enabling real-time interaction among multiple human drivers. High-resolution trajectory data were collected from 231 valid interaction events. An eight-direction relative position model was employed to classify behaviors into four patterns: longitudinal, lateral, front cut-in, and rear cut-in. Risk was quantified using time-exposed and time-integrated Anticipated Collision Time metrics, with events subsequently clustered into low (n = 138), medium (n = 67), and high-risk (n = 26) categories. An ordered logit regression model identified key risk factors. The results quantitatively demonstrate that interaction risk escalates significantly with abrupt speed changes (OR = 16.22) and late-stage occurrence of speed extremes (OR = 6.76) in the interacting vehicle, as well as large initial speed differences (OR = 2.45). Conversely, stable speed regulation and adaptive acceleration by the subject vehicle proved to be potent mitigating factors. These findings provide actionable insights for the development of intelligent collision warning systems and the sustainable design of interchange infrastructure. Full article

High-accuracy spatiotemporal monitoring of surface nitrogen dioxide (NO₂) concentrations is essential for air quality management. This study evaluates machine learning-based estimates of near-surface NO₂ concentrations using data from the geostationary GEMS instrument and the polar-orbiting TROPOMI over China in 2022. Four tree-based models—Random Forest, XGBoost, CatBoost, and LightGBM—were trained by integrating satellite vertical-column densities with multi-source meteorological and ancillary data. Results show that CatBoost achieved the highest accuracy, with an R² of 0.842 for GEMS and 0.765 for TROPOMI, alongside the lowest RMSE and MAE. Models trained on GEMS data consistently outperformed TROPOMI-based models across all metrics. This advantage is primarily attributed to the substantially larger training sample size enabled by GEMS’s high temporal resolution, as confirmed through a controlled experiment with consistent sample sizes which isolated the effect of data volume. Spatially, GEMS estimates captured sharper concentration gradients and localized emission hotspots, while TROPOMI produced smoother fields. Temporally, only GEMS allowed the reconstruction of detailed diurnal patterns and near-real-time pollution episode tracking. This study confirms the significant added value of geostationary satellite data for high-frequency air quality monitoring and analysis when combined with machine learning. Full article