Search Results (285)

Accurate rainfall monitoring is vital for hydrology and environmental sensing. This study presents a physics-aware deep learning framework using E-band (71–86 GHz) commercial microwave links (CMLs). Using the extensive urban CML dataset and methodology, a bi-directional Long Short-Term Memory (Bi-LSTM) model is developed to classify wet and dry periods under a temporal generalization framework across heterogeneous link configurations. The approach integrates physical signal decomposition, including baseline estimation, gaseous attenuation correction, and wet antenna attenuation (WAA) modeling, with sequence-based learning. Results demonstrate that the temporal deep learning model outperforms classical threshold-based and physical k–R approaches when evaluated over independent temporal validation blocks, effectively reducing sensitivity to path-length-related variability on heterogeneous paths. The model maintains stable performance (loss < 3%) under moderate signal-level noise. SHapley Additive exPlanations (SHAP) confirm the model relies on physical features, such as signal volatility and temporal trends, to reliably differentiate rainfall from WAA. This framework highlights the potential of E-band infrastructure as a distributed sensing network for integrated sensing and communication (ISAC) architectures. Full article

With the growing focus on people-centered urban development sustainability in megacities, urban amenities have emerged as an important factor consistently associated with residential differentiation and restructuring. Understanding how it relates to the structure of social space is essential to advancing spatial equity. The study developed an analytical framework that integrates functional characteristics and supply patterns and applied Multi-scale Geographically Weighted Regression (MGWR) to examine how amenities shaped socio-spatial differentiation within Beijing’s Fifth Ring Road from 2015 to 2025. The results indicate that socio-spatial differentiation showed a rise followed by a decline across the three time points examined, yet its spatial pattern maintained a stable agglomeration characteristic of “high in the core area and low in the peripheral areas.” Significant differences exist in the roles of amenities across different attributes and scales. Market-driven factors, represented by amenity density and amenity diversity, typically exert their influence over larger spatial scales and are generally associated with spatial mixing and provide baseline opportunities for potential social interaction. Attributes such as amenity publicness and amenity uniqueness, which are more influenced by institutional and capital factors, primarily operate at local scales. While they are often associated with exclusionary effects in traditional core areas, they are also consistent with a certain degree of spatial integration in some revitalized districts. This study offers a more nuanced explanation for understanding the socio-spatial restructuring of megacities in transition and provides empirical evidence for advancing more equitable and sustainable urban governance. Full article

The increasing integration of multimodal technologies, including augmented and virtual reality and interactive digital systems, has shifted the focus of innovation from technological capability to user experience and interaction. In organizational settings, leadership structures play a key role in shaping how these technologies are designed and experienced. This study examines how leadership control orientation influences innovation outcomes through multimodal experience design and cognitive burden. Using structural equation modeling on a sample of 3017 employees who actively use multimodal systems, the study develops a process-based model linking leadership, multimodal experience design, cognitive burden, and innovation. The findings suggest that control-oriented leadership is negatively associated with multimodal experience design and positively associated with cognitive burden, whereas well-structured multimodal systems are associated with lower levels of cognitive burden. Multimodal design emerges as a central driver of perceived innovation, whereas cognitive overload negatively affects innovation outcomes. The results further reveal a sequential mediation process involving multimodal experience design and cognitive burden. Multi-group analysis confirms that these relationships remain stable across different levels of environmental control. The study contributes by integrating leadership, human–technology interaction, and experience design into a unified framework, offering a process-oriented explanation of innovation in multimodal environments. Full article

Accurate viral protein function annotation is essential for genomic surveillance, yet conventional retrieval-augmented generation (RAG) pipelines often fragment biological evidence into fixed-length text chunks, disrupting relationships among ORFs, annotations, structural domains, sequence motifs, residue mappings, and model-derived attention evidence. We propose ViroBioTree, a tree-structured biological evidence retrieval framework for downstream viral protein evidence review rather than a new primary annotation classifier. Built as an evidence organization layer on ViralMultiNet-derived ORF-level predictions and annotations, ViroBioTree converts sequence, annotation, structure, and attention evidence into typed biological nodes and traceable edges, then performs deterministic multi-channel recall, evidence-aware reranking, balanced TopK selection, rule-based verification, and node-cited report generation. In a demo benchmark, ViroBioTree achieved its strongest deterministic proxy performance on structure-explanation tasks, with Precision@K = 1.0, Recall@K = 1.0, and diversity = 0.52; these values reflect expected node-type and tag agreement rather than independent biological correctness. A bounded full-scale SARS-CoV-2 index contained 39,800 ORF rows, 80,000 attention records, 199,418 nodes, and 495,886 edges. In a stratified full20k diagnostic evaluation, ViroBioTree showed task-dependent advantages over LlamaIndex vector retrieval for conflict detection, evidence retrieval, and structure explanation, while LlamaIndex remained competitive or stronger for annotation-rich function annotation. A cross-family Influenza A Virus (IAV) diagnostic audit showed that the schema can represent IAV evidence namespaces while explicitly exposing missing formal ORF inputs, missing attention evidence, and unavailable residue/PDB assertions. Supplementary robustness, external sanity-check, diversity-risk, expert-evaluation, domain-tool positioning, and cross-family audit analyses supported traceability, report quality, and conservative evidence handling, but also showed that stable Precision@K under query perturbation does not necessarily imply stable retrieved evidence sets. ViroBioTree operates offline and deterministically, but does not address raw-read assembly, base calling, primary ORF prediction, or wet-lab validation. Its results should be interpreted as proxy and expert-reviewed evidence for traceable viral protein evidence retrieval and report generation rather than as direct validation of biological function annotation. Full article

This study examines customer retention intention in neobanking environments using a theory-informed explainable machine learning framework. Existing digital banking research typically relies on linear modelling approaches to explain retention behaviour, potentially overlooking nonlinear, value-range-dependent, and interaction-based predictive patterns. Using a publicly available survey of 305 neobank users, this study compares regularized linear models, a partial least squares structural equation modelling (PLS-SEM)-inspired benchmark, and XGBoost (version 3.2.0) under repeated nested cross-validation. SHapley Additive exPlanations (SHAP)-based explainability, SHAP interaction analysis, generalized additive model (GAM) diagnostics, construct-level aggregation, and construct-sensitivity checks are used to interpret model behaviour and assess robustness. The results show that XGBoost substantially outperforms the linear benchmarks, achieving the lowest average RMSE and highest average R² across 100 out-of-sample test-fold estimates. Trust-related indicators provide the largest share of model-based predictive importance, followed by perceived security and switching costs. SHAP and GAM diagnostics suggest that trust and switching costs may contribute to retention intention in heterogeneous and nonlinear ways, while perceived security displays a more stable positive predictive pattern. Age-related nonlinearities appear weak and should be interpreted cautiously given the young sample profile. The analysis also suggests possible non-additive relationships between trust and perceived security. The study contributes to digital banking and FinTech research by showing how explainable machine learning can complement theory-driven retention models, identify potentially nonlinear predictive patterns, and preserve interpretability. The findings offer practical insight for trust-building, visible security assurance, and retention diagnostics in neobanking contexts. Full article

Sepsis is life-threatening organ dysfunction caused by a dysregulated host response to infection, making early detection critical for improving outcomes in intensive care units (ICUs). This study presents a retrospective comparative evaluation of deep learning architectures for predicting sepsis up to 6 h before the PhysioNet/Computing in Cardiology 2019 Challenge onset label using hourly structured electronic health record (EHR) variables, including vital signs, laboratory measurements, and demographics. Evaluated architectures include Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional Long Short-Term Memory (Bi-LSTM), Temporal Convolutional Network (TCN), Transformer, and hybrid Convolutional Neural Network–Vision Transformer (CNN-ViT) models. Median imputation and class-weighted loss were applied to address missing values and severe class imbalance, while Shapley Additive Explanations (SHAP) and attention analyses were used as complementary interpretability approaches. Among the evaluated models, CNN-ViT achieved the strongest overall minority-class performance, with 88.25% accuracy, 0.7480 recall, a 0.454 F1-score, and a 0.48 area under the precision–recall curve (AUPRC), although the numerical gains over other advanced temporal and hybrid architectures were modest. Leave-one-unit-out evaluation further demonstrated relatively stable performance under internal distribution shifts. The results suggest that combining local feature extraction with temporal and attention-based modeling can improve early sepsis prediction from structured ICU data. However, the study represents a retrospective computational benchmark using a public dataset and does not constitute prospective clinical validation or real-world deployment assessment. Full article

Background: Accurate survival prediction is essential for optimizing treatment planning in patients with castration-resistant prostate cancer (CRPC). However, traditional statistical models often underperform because of limited variable inclusion and an inability to account for complex, multidimensional data interactions. Methods: We retrospectively collected 46 clinical, laboratory, and pathological variables from 801 patients with CRPC, covering the disease course from initial diagnosis to CRPC progression. Multiple machine learning (ML) models, including random survival forests (RSF), XGBoost, LightGBM, and logistic regression, were developed to predict cancer-specific mortality (CSM), overall mortality (OM), and 2- and 3-year survival status. The dataset was divided into training and test cohorts (80:20), and 10-fold cross-validation was performed. Performance was assessed using the C-index for regression models and the area under the curve (AUC), accuracy, precision, recall, and F1-score for classification models. Model interpretability was evaluated using SHapley Additive exPlanations (SHAP). Results: Over a median follow-up of 24 months, 70.6% of patients experienced CSM. Although XGBoost with its own imputation method achieved the highest C-index in the validation set, RSF demonstrated more stable performance and achieved the highest C-index in the held-out test set for both CSM (0.772) and OM (0.771). For classification tasks, RSF demonstrated superior performance in predicting 2-year survival, whereas XGBoost achieved the highest F1-score for 3-year survival prediction. SHAP analysis identified time to first-line CRPC treatment, hemoglobin level, and alkaline phosphatase level as key predictors of survival outcomes. Conclusions: RSF demonstrated robust test-set performance for time-to-event prediction, whereas XGBoost showed complementary value for 3-year survival classification. These models provide accurate and interpretable prognostic tools that may support personalized treatment strategies. External validation and integration of emerging therapies are warranted to enhance broader clinical applicability. Full article

Background/Objectives: The aim of this study was to develop and validate a high-frequency ultrasound radiomics-based model for quantitative assessment of metacarpophalangeal (MCP) joint cartilage damage in early rheumatoid arthritis (RA). Methods: 656 MCP joints from 99 early RA patients and 65 healthy controls were prospectively enrolled and graded according to the Outcome Measures in Rheumatology (OMERACT) system. After radiomics feature extraction, five machine learning classifiers were evaluated. Radiomics, clinical, and combined models were constructed and assessed. Radiomics scores were compared among healthy grade 0 joints, early RA grade 0 joints stratified into two risk subgroups, and RA grade ≥ 1 joints. SHapley Additive exPlanations (SHAP) analysis was used for interpretation. Results: Eight stable radiomics features were selected. Among classifiers, support vector machine achieved the highest cross-validated performance and was selected as the final radiomics classifier (validation AUC = 0.804). The combined model, integrating radiomics features with age, disease duration, and Disease Activity Score in 28 joints, achieved the best diagnostic performance (AUC = 0.855), significantly outperforming both the radiomics and clinical models. Among OMERACT grade 0 joints, the high-risk subgroup demonstrated elevated radiomics-derived scores. SHAP analysis identified original_shape2D_PerimeterSurfaceRatio as the strongest contributor. Conclusions: High-frequency ultrasound radiomics combined with clinical features demonstrated strong performance in detecting MCP joint cartilage damage in early RA and may provide a quantitative extension to conventional semiquantitative assessment. Full article

Background: Spinal deformities in children represent a relevant public health issue, with possible long-term consequences. Timely identification of their determinants is essential for adequate prevention. Methods: This study was a secondary analysis of data from the 2019 Serbian National Health Survey, including 1309 children aged 5–14 years. Logistic regression with LASSO regularization and multiple ML algorithms were tested, with XGBoost selected as the optimal model. Class imbalance was addressed using class weighting and SMOTE. Model interpretability was achieved using SHAP analysis. Results: The prevalence of spinal deformities was 8.6%. Univariable analyses showed that age, poorer self-rated health, chronic illness, recent injuries, and pes planus were significantly associated with spinal deformities. Family-related variables showed no significant associations. Among the evaluated models, XGBoost demonstrated the most stable performance across the applied evaluation metrics and the best balance between predictive performance and interpretability. SHapley Additive exPlanations (SHAP) analysis showed that pes planus was the strongest determinant, followed by age and chronic illness, while socio-demographic and family factors had minimal influence. Conclusion: Explainable machine learning models, particularly XGBoost combined with SHAP, can allow for the identification and interpretation of key determinants of spinal deformities in children. Pes planus was shown to be modifiable and relevant associated determinant, supporting its importance in early screening and preventive strategies. Full article

Cropland non-grain use has become an important challenge for food security and cropland governance in rapidly urbanising agricultural regions, yet its trajectory heterogeneity and the divergence between current spatial patterns and long-term-change mechanisms remain insufficiently understood. Taking the Changsha–Zhuzhou–Xiangtan (CZT) urban agglomeration in China as a case, this study quantified the cropland non-grain rate (NGR) on a 1 km grid for 2000, 2010, and 2020, classified grid-level transition trajectories, and developed three temporally structured eXtreme Gradient Boosting (XGBoost) models with spatial block cross-validation, Shapley additive explanations (SHAP) interpretation, and geographically explicit SHAP (GeoSHAP) local attribution. The results show that low-NGR and stable grids formed the dominant regional background, while recent NGR increases were mainly concentrated along the urban development corridor and metropolitan fringe. Current NGR status and long-term NGR change showed divergent explanatory structures. The current spatial pattern was mainly associated with terrain constraints and contemporary urban pressure, whereas long-term change was more strongly conditioned by baseline urbanisation and subsequent urban–environmental changes. Nonlinear dependence analysis further identified model-derived response zones related to slope, impervious surface conditions, hydrothermal change, and hydrological proximity. GeoSHAP mapping revealed that locally dominant mechanisms varied substantially across the study area, indicating that cropland non-grain use was shaped by spatially heterogeneous combinations of terrain, urbanisation, hydrothermal background, and hydrological context. These findings support a shift from aggregate status monitoring toward trajectory-specific and mechanism-differentiated cropland management in urban agglomerations. Full article

Soil organic carbon (SOC) of cropland is a key indicator of soil fertility and contributes to climate regulation and carbon storage. The understanding of SOCchanges in cropland in Northeast China still lacks high-precision long-term empirical evidence. This study is of great significance for ensuring national food security and regional sustainable development. Taking Changchun, a representative black soil region, as the study area, this study integrated 953 field samples with 19 predictors to estimate cropland soil organic carbon density (SOCD) from 2000 to 2022. The performance of quantile regression neural network (QRNN), random forest (RF), and extreme gradient boosting (XGBoost) models was compared. QRNN showed the best overall performance (R² = 0.74, RMSE = 0.57 kg/m², MAE = 0.40 kg/m², and RPIQ = 2.46) and also exhibited greater stability in temporal-stage validation. Results indicated that SOCD exhibited an overall declining trend with intermittent recoveries, decreasing from 3.72 kg/m² in 2000 to 3.36 kg/m² in 2005, then increasing to 3.55 kg/m² in 2010, slightly declining to 3.46 kg/m² in 2015, and recovering to 3.63 kg/m² in 2022. Spatially, SOCD remained low in the southwest, fluctuated markedly in the north, and was relatively stable in the central region. The analysis of the optimal parameter geographic detector (OPGD) showed that Y-latitude, elevation, and mean annual temperature (MAT) were stable dominant factors, while precipitation (PRE) and remote sensing variables showed stage-dependent effects. Interactions among multiple factors further enhanced the explanation of SOCD variations. These findings provide theoretical support for enhancing soil carbon retention and promoting long-term cropland sustainability in black soil areas. Full article

This study investigated the effects of hydrolysed fish and bovine collagens at 1.25%, 2.50%, and 5.00% on the fermentation kinetics, physicochemical quality, and refrigerated storage stability of fermented milk beverages enriched with vitamin C. The work addressed three linked questions: whether collagen source determines the technological response of the dairy matrix, whether these effects are dose-dependent, and whether the observed changes remain relevant during 28 days of storage at 6 °C. During fermentation at 37 °C, 1.25% fish collagen maintained acidification kinetics comparable to the control, whereas bovine collagen, especially at higher doses, prolonged the time required to reach pH 4.6. During storage, fish collagen demonstrated better technological compatibility with the fermented milk matrix, improving water-holding capacity and maintaining or increasing gel hardness, whereas bovine collagen weakened the gel structure but showed a stronger buffering effect and higher pH values. Starter culture viability was maintained throughout storage: S. thermophilus remained highly stable, whereas Lactobacillus spp. declined gradually to approximately 5.7–6.1 log CFU/g by day 28. Colour analysis showed a progressive increase in yellowness (b*) and total colour difference (ΔE*) in all samples, with the magnitude depending on collagen source, dose, and storage time. This study indicates that hydrolysed fish collagen is generally more compatible with fermented milk enriched with vitamin C when structural stability and water retention are prioritised. However, batch-specific molecular-weight distribution and amino acid composition of the hydrolysates were not determined; therefore, source-related mechanisms are interpreted as plausible technological explanations rather than direct molecular evidence. Full article

Understanding the complex relationship between the built environment and urban vitality is essential for evidence-based urban renewal. However, most existing studies rely on linear regression models that fail to capture the non-linear threshold effects inherent in urban systems and depend on costly proprietary datasets that limit reproducibility. This study proposes a scalable, open-data-driven framework to decode the non-linear mechanisms governing population-based urban vitality in Zhengzhou, a rapidly regenerating metropolis in Central China. Using Areas of Interest (AOIs) as functional spatial units to mitigate the Modifiable Areal Unit Problem (MAUP), we construct a multidimensional built environment indicator system (5D+S: Density, Diversity, Design, Distance to Transit, Destination Accessibility, and Surroundings) from multi-source open data, including 100 m WorldPop population grids, OpenStreetMap building vectors, Points of Interest (POIs), and transit station data. An explainable machine learning approach combining XGBoost with SHapley Additive exPlanations (SHAP) is employed to identify the relative importance of built environment factors and quantify their non-linear threshold effects on population-based urban vitality (operationally defined as residential population density derived from WorldPop 100 m grids). Across 3920 AOIs, XGBoost (R² = 0.846, RMSE = 0.104) substantially outperforms Ordinary Least Squares regression (R² = 0.634), confirming pervasive non-linear relationships, with stable 5-fold cross-validated R² = 0.713 ± 0.115. SHAP analysis reveals four dominant drivers: Distance to Commercial Core (DistCBD), Bus Station Density within 500 m (BusDen500), Green Coverage Ratio (GreenRatio), and Building Density (BD). Critical thresholds are identified: vitality contributions decay sharply beyond approximately 4.3 km from the CBD; at least 4 bus stations within 500 m are required for meaningful transit benefit; building density delivers positive returns within a 2–30% range; and excessive green coverage above 8.5% within 500 m is associated with declining population-based vitality, a finding that reflects spatial competition between ecological land use and residential density rather than a negative effect of greenery per se. These findings provide quantitative design guidelines for precision urban renewal, moving beyond “the more, the better” planning assumptions to identify optimal intervention ranges. Full article

This study is set against a backdrop of interest in understanding how secondary school pupils perceive learning experiences based on Augmented Reality (AR) and digital assessment, with the aim of designing a useful tool for comparative and replicable research in AR-STEM contexts. To this end, an attitudinal questionnaire was developed and preliminarily validated through a quantitative study conducted with a sample of 199 students from various schools who worked on science curriculum content using an AR application created and validated for the explanation of key concepts. The instrument, designed ad hoc based on the TAM and IMMS models and peer-reviewed, comprised 35 items. Its reliability and construct validity were analysed using Cronbach’s alpha and exploratory and confirmatory factor analysis. The results showed favourable scores for motivation, technological acceptance and the evaluation of digital assessment, alongside high internal consistency and a stable three-factor structure. In conclusion, the instrument presented in full in this study provides preliminary evidence that it is a valid, reliable and useful tool, although the digital assessment dimension remains emerging and will require further validation with independent samples. Full article

Background: Mechanistic reporting from large-scale biomedical knowledge graphs remains challenging, particularly when integrating structured graph evidence with large language model (LLM)–based explanation in a reproducible and auditable manner. Existing approaches either rely on manual synthesis of graph-derived results or generate unconstrained narratives that lack traceability to underlying evidence. Methods: We present SemNet Explorer, an evidence-grounded knowledge graph–LLM unified framework for automated mechanistic reporting across biomedical domains using SemNet 2.0, a PubMed-scale heterogeneous knowledge graph. Given a set of target concepts and a selected semantic layer, the framework organizes graph-derived evidence into structured regions and generates two complementary report types: global reports for process-level mechanisms and anchor-centric reports for localized mediator-based explanations. A central methodological contribution is an ablation-derived adaptive grounding policy: we systematically compare alternative evidence-integration strategies across report types, semantic layers, and region structures, and use the resulting preferences to guide prompt selection in the deployed system. Results: SemNet Explorer produces stable region decompositions and interpretable report scaffolds across molecular (AAPP), disease-level (DSYN), and pharmacologic (PHSU) representations. For global reports, explicit evidence grounding improves expression quality more consistently than content accuracy, with benefits dependent on evidence density and semantic abstraction. In contrast, anchor-centric reports show consistent improvements in both content and expression under stronger, mediator-constrained prompting. These findings are supported by both pairwise ablation comparisons and absolute score analyses. Conclusions: SemNet Explorer establishes a generalizable unified framework and interactive platform for transforming knowledge graph evidence into reproducible mechanistic narratives across biomedical domains, including multimorbidity analysis, comparative pathophysiology, drug repurposing, and adverse event discovery. The results demonstrate that effective knowledge graph–LLM integration requires adaptive, context-dependent evidence grounding rather than fixed prompting strategies. Full article