Search Results (23,517)

Background/Objectives: Accurate prognostication for patients with chronic critical illness (CCI) following brain injury remains challenging. Conventional scoring systems like the Acute Physiology and Chronic Health Evaluation II (APACHE II) and the Nutrition Risk in the Critically Ill (NUTRIC) score are validated as “left-side” models for risk stratification at intensive care unit (ICU) admission but may not capture the evolving trajectory of prolonged illness. This study aimed to evaluate the prognostic performance of APACHE II and NUTRIC as “right-side” models—assessed at intervals closer to the outcome—by testing the hypothesis that their predictive accuracy for in-hospital mortality improves when measured nearer to the time of death. Methods: In this real-world data analysis study, data were extracted from the electronic health records (Russian Intensive Care Dataset [RICD] v. 2.0) of 328 adult patients with CCI following brain injury. The discriminative ability of repeatedly assessed APACHE II and NUTRIC scores for predicting mortality was analyzed by calculating the area under the receiver operating characteristic curve (AUROC) for three predefined intervals before death: within ≤7 days, 8–14 days, and ≥15 days. Results: Among the 328 patients (median age 64 years; 18.3% in-hospital mortality), a total of 380 paired score assessments were analyzed. The predictive performance for both scores was highest within 7 days of death (APACHE II AUROC: 0.883; NUTRIC AUROC: 0.839). Discriminatory ability declined at 8–14 days (APACHE II AUROC: 0.807; NUTRIC AUROC: 0.778) and was poorest at ≥15 days before death (APACHE II AUROC: 0.671; NUTRIC AUROC: 0.681). The NUTRIC score consistently demonstrated higher AUROC values than APACHE II across all intervals, though the differences were not statistically significant. Conclusions: In patients with CCI following brain injury, the prognostic accuracy of APACHE II and NUTRIC scores is time-dependent, peaking immediately before death and offering poor long-term prediction from admission. These findings underscore the limitation of static, admission-based models and highlight the necessity for developing dynamic, personalized and time-sensitive prognostic tools tailored to the evolving course of chronic critical illness. Full article

Identifying abnormal group behavior formed by multi-type participants from large-scale historical industry and tax data is important for regulators to prevent potential criminal activity. We propose an Abnormal Alliance detection framework comprising two methods. For detecting joint behavior among multi-type participants, we present DyCIAComDet, a dynamic community identification and tracking method for large-scale, time-varying bipartite multi-type participant networks, and introduce three community-splitting measurement indicators—cohesion, integration, and leadership—to improve community division. To verify whether joint behavior is abnormal, termed an Abnormal Alliance, we propose BMPS, a frequent-sequence identification algorithm that mines key features along community evolution paths based on bitmap matrices, sequence matrices, prefix-projection matrices, and repeated-projection matrices. The framework is designed to address sampling limitations, temporal issues, and subjectivity that hinder traditional analyses and to remain scalable to large datasets. Experiments on the Southern Women benchmark and a real tax dataset show DyCIAComDet yields a mean modularity Q improvement of 24.6% over traditional community detection algorithms. Compared with PrefixSpan, BMPS improves mean time and space efficiency by up to 34.8% and 35.3%, respectively. Together, DyCIAComDet and BMPS constitute an effective, scalable detection pipeline for identifying abnormal alliances in tax datasets and supporting regulatory analysis. Full article

Local pier scour remains one of the leading causes of bridge failure, calling for predictions that are both accurate and uncertainty-aware. This study develops an interpretable data-driven framework that couples CatBoost (Categorial Gradient Boosting) for deterministic point prediction with NGBoost (Natural Gradient Boosting) for probabilistic prediction. Both models are trained on a laboratory dataset of 552 measurements of local scour at bridge piers using non-dimensional inputs (y/b, V/V_c, b/d₅₀, Fr). Model performance was quantitatively evaluated using standard regression metrics, and interpretability was provided through SHAP (Shapley Additive Explanations) analysis. Monte Carlo–based reliability analysis linked the predicted scour depths to a reliability index β and exceedance probability through a simple multiplicative correction factor. On the held-out test set, CatBoost offers slightly higher point-prediction accuracy, while NGBoost yields well-calibrated prediction intervals with empirical coverages close to the nominal 68% and 95% levels. This framework delivers accurate, interpretable, and uncertainty-aware scour estimates for target-reliability, risk-informed bridge design. Full article

Predicting the degradation and lifespan of embedded computers relies critically on the accurate evaluation of key parameter degradation within computing systems. Accelerated high-temperature tests are frequently employed as an alternative to ambient-temperature degradation tests, primarily due to the excessive duration and cost of ambient-temperature testing. However, the scarcity of effective methodologies for correlating degradation trends across distinct temperature conditions persists as a prominent challenge. This study addresses this gap by leveraging adversarial learning to generate low-temperature degradation sequences from high-temperature datasets. The adversarial learning framework enables feature transfer across diverse operating conditions and facilitates domain adaptation learning. This empowers the model to extract features invariant to degradation trends across multiple temperature conditions. Furthermore, soft dynamic time warping (SDTW) is utilized to precisely align the generated low-temperature sequences with their real-world counterparts. This alignment methodology enables elastic matching of time series data exhibiting nonlinear temporal variations, thereby ensuring accurate comparison and synchronization of degradation sequences. Compared with prior methodologies, our proposed approach delivers superior performance on computer degradation data. It offers a more accurate and reliable solution for the degradation analysis and lifespan prediction of embedded computers, thereby advancing the reliability of computational systems. Full article

Background: Loneliness (LON) is a heritable psychosocial trait that frequently co-occurs with major depressive disorder (MDD) and sleep traits. Despite known genetic contributions, the shared genetic architecture and molecular mechanisms underlying their co-occurrence remain largely unknown. This study aimed to uncover novel genetic risk loci and cross-trait gene expression effects. Methods: Large-scale genome-wide association study (GWAS) datasets were analyzed using the causal mixture model (MiXeR) to estimate polygenicity and shared genetic architecture. Genetic correlation analyses were performed using linkage disequilibrium score regression (LDSC) and local analysis of [co]variant annotation (LAVA). Conditional and conjunctional FDR methods further identified single nucleotide polymorphisms (SNPs). FUMA was used for gene mapping and annotation, and transcriptome-wide association studies (TWAS) assessed cross-trait gene expression effects. Results: Analyses revealed extensive polygenic overlap between LON, MDD, and sleep-related traits, with concordant and discordant effects. Several novel loci were identified, and cross-trait gene expression effects were observed in multiple brain-expressed genes, including WNT3, ARHGAP27, PLEKHM1, and FOXP2. These findings provide insight into the shared genetic architecture and relevance of these traits. Conclusions: This study demonstrates a significant shared polygenic architecture among LON, MDD, and sleep traits, providing new biological insights. It advances our understanding of cross-trait genetic mechanisms and identifies potential targets for future research, offering broader implications for trait co-occurrence. Full article

As encrypted traffic analysis becomes increasingly vital for network security, the conventional reliance on centralized classification faces growing challenges due to data privacy regulations and data silos across heterogeneous nodes. Federated learning (FL) emerges as a solution by training models locally and sharing only parameter updates, thus preserving privacy. However, its performance is significantly degraded by data heterogeneity (i.e., non-IID data) among participants. To address this critical challenge, this paper proposes a Federated Learning framework based on Deep Mutual Learning (FLDML). In this method, clients first train local models on their private traffic data and then upload them to a server. There, they engage in deep mutual learning through co-training on a shared public dataset to enhance robustness and mitigate data heterogeneity. Subsequently, a global classifier is generated by averaging the model parameters. When evaluated on the ISCX VPN-NonVPN 2016 dataset, FLDML demonstrates significantly superior performance in handling non-IID traffic data compared to classical FL algorithms. This study concludes that the proposed framework not only effectively mitigates data heterogeneity in federated scenarios but also provides a scalable and improved solution for distributed network traffic classification. Full article

This study investigated the statistical association between public incentives and industrial innovation as reflected in firms’ financial performances. In particular, the analysis was carried out considering a Regional Operational Program, namely, the 2007–2013 ERDF Regional Program in Lombardy, and investigating a dataset of Lombardy-based companies that received support through the mentioned initiative. For each of them, balance sheet variables before and after the acquisition of the incentive and the development of the related innovation project were detected and analyzed by means of both standard and normalized linear regression. Notably, normalized regressions showed that higher subsidy intensity was positively associated with subsequent changes in revenues and intangible assets, especially among manufacturing firms, thereby supporting policies that target sectors with a high innovation capacity. Furthermore, this research underscores the importance of tailoring policy instruments to local and sectoral contexts, recognizing the limitations of one-size-fits-all approaches. In keeping with this exploratory stance, this study does not build a counterfactual control group and makes no causal claims; it simply documents balance sheet associations that may inform future, impact-oriented research. Given the absence of a control group, the design is observational; all findings describe associations and do not allow causal inference. Full article

The Tensift River Basin, part of the Mediterranean region, faces significant agricultural losses due to increasing drought frequency and severity, impacting up to 15% of the national GDP. The increasing climate crisis demands our immediate attention and proactive adaptation measures, including the enhancement of early-warning tools to support timely and informed responses. To this end, our study aims to achieve the following goals: (1) evaluate satellite and reanalysis products against in situ observations using statistical metrics; (2) identify the best probability distribution for calculating drought indices using goodness-of-fit testing; (3) compare the performances of the Standardized Precipitation Index (SPI) and the Standardized Precipitation and Evapotranspiration Index (SPEI) at different aggregation timescales by comparing index-based and reported (i.e., impact-based) drought events using receiver operating characteristic (ROC) analysis. Our findings indicate that CHIRPS and ERA5-Land datasets perform well compared to in situ measurements for drought monitoring in the Tensift River Basin. Pearson Type 3 was identified as the optimal distribution for SPI calculation, while log-logistic was confirmed for SPEI. We also explored the effect of using the Thornthwaite method and the Hargreaves method when computing the SPEI. These results can serve as a basis for drought monitoring, modeling, and forecasting, to support decision-makers in the sustainable management of water resources. Full article

Accurate mapping of burned forest areas in mountainous regions is essential for wildfire assessment and post-fire ecological management. This study develops an FS-SNIC-ML workflow that integrates multi-source optical fusion, semi-automatic sample generation, feature selection, and object-based machine-learning classification to support reliable burned-area mapping under complex terrain conditions. A pseudo-invariant feature (PIFS)-based fusion of Sentinel-2 and Landsat 8 imagery was employed to generate cloud-free, gap-free, and spectrally consistent pre- and post-fire reflectance datasets. Burned and unburned samples were constructed using a semi-automatic SAM–GLCM–PCA–Otsu procedure and county-level stratified sampling to ensure spatial representa-tiveness. Feature selection using LR, RF, and Boruta identified dNBR, dNDVI, and dEVI as the most discriminative variables. Within the SNIC-supported GEOBIA framework, four classifiers were evaluated; RF performed best, achieving overall accuracies of 92.02% for burned areas and 94.04% for unburned areas, outperforming SVM, CART, and KNN. K-means clustering of dNBR revealed spatial variation in fire conditions, while geographical detector analysis showed that NDVI, temperature, soil moisture, and their pairwise interactions were the dominant drivers of wildfire hotspot density. The proposed workflow provides an effective and transferable approach for high-precision burned-area extraction and quantification of wildfire-driving factors in mountainous forest regions. Full article

The rapid expansion of Internet of Things (IoT) technologies has introduced significant challenges in understanding the complexity and structure of network traffic data, which is essential for developing effective cybersecurity solutions. This research presents a comprehensive statistical and multivariate analysis of the IoT-23 dataset to identify meaningful network traffic patterns and assess the effectiveness of various analytical methods for IoT security research. The study applies descriptive statistics, inferential analysis, and multivariate techniques, including Principal Component Analysis (PCA), DBSCAN clustering, and factor analysis (FA), to the publicly available IoT-23 dataset. Descriptive analysis reveals clear evidence of non-normal distributions: for example, the features src_bytes, dst_bytes, and src_pkts have skewness values of −4.21, −3.87, and −2.98, and kurtosis values of 38.45, 29.67, and 18.23, respectively. These values indicate highly skewed, heavy-tailed distributions with frequent outliers. Correlation analysis revealed a strong positive correlation (0.97) between orig_bytes and resp_bytes, and a strong negative correlation (−0.76) between duration and resp_bytes, while inferential statistics indicate that linear regression provides optimal modeling of data relationships. Key findings show that PCA is highly effective, capturing 99% of the dataset’s variance and enabling significant dimensionality reduction. DBSCAN clustering identifies six distinct clusters, highlighting diverse network traffic behaviors within IoT environments. In contrast, FA explains only 11.63% of the variance, indicating limited suitability for this dataset. These results establish important benchmarks for future IoT cybersecurity research and demonstrate the superior effectiveness of PCA and DBSCAN for analyzing complex IoT network traffic data. The findings offer practical guidance for researchers in selecting appropriate statistical methods for IoT dataset analysis, ultimately supporting the development of more robust cybersecurity solutions. Full article

The increasing use of blockchain smart contracts has introduced new security challenges, as small coding errors can lead to major financial losses. While rule-based static analyzers remain the most common detection tools, their limited adaptability often results in false positives and outdated vulnerability patterns. This study presents a comprehensive comparative analysis of machine learning (ML) and deep learning (DL) methods for smart contract vulnerability detection using the BCCC-SCsVuls-2024 benchmark dataset. Six models (Random Forest, k-Nearest Neighbors, Simple and Deep Multilayer Perceptron, and Simple and Deep one-dimensional Convolutional Neural Networks) were evaluated under a unified experimental framework combining RobustScaler normalization and Principal Component Analysis (PCA) for dimensionality reduction. Our experimental results from a five-fold cross-validation show that the Random Forest classifier achieved the best overall performance with an accuracy of 89.44% and an F1-score of 93.20%, outperforming both traditional and neural models in stability and generalization. PCA-based feature analysis revealed that opcode-level features, particularly stack and memory manipulation instructions (PUSH, DUP, SWAP, and RETURNDATASIZE), were the most influential in defining contract behavior. Full article

Biological samples of non-human origin, commonly encountered in wildlife crime investigations, present distinct challenges regarding forensic DNA analysis efforts. Although the types of samples encountered in human identity testing can vary to some degree, analyzing DNA from one species is facilitated by unified processes, common genetic marker systems, and national DNA databases. In contrast, non-human animal species identification is confounded by a diverse range of target species and a variety of sampling materials, such as feathers, processed animal parts in traditional medicine, and taxidermy specimens, which often contain degraded DNA in low quantities, are contaminated with chemical inhibitors, and may be comingled with other species. These complexities require specialized analytical approaches. Compounding these issues is a lack of validated non-human species forensic sampling and typing kits, and the risk of human DNA contamination during evidence collection. Markers residing on the mitochondrial genome (mtDNA) are routinely sought because of the large datasets available for comparison and their greater sensitivity of detection. However, the barcoding results can be complicated at times for achieving species-level resolution, the presence of nuclear inserts of mitochondrial DNA (NUMTs), and the limitation of mtDNA analysis alone to detect hybrids. Species-specific genetic markers for identification have been developed for a few high-profile species; however, many CITES (Convention on International Trade in Endangered Species of Wild Fauna and Flora)-listed organisms lack specific, validated forensic analytical tools, creating a significant gap in investigative enforcement capabilities. This deficiency stems in part from the low commercial nature of wildlife forensics efforts, a government research-driven field, the difficulty of obtaining sufficient reference samples from wild populations, limited training and education infrastructure, and inadequate funding support. Full article

X-ray absorption near-edge structure (XANES) spectra are employed to characterise the coordination numbers of metallic elements within materials. However, conventional XANES analysis methods frequently rely on preconceived assumptions regarding the analysed samples, which may not fully satisfy the requirements of scientific research and industrial applications. To mitigate such reliance, a novel approach based on the Gated Adaptive Network for Deep Automated Learning of Features (GANDALF) is proposed. To effectively extract multi-scale information from the XANES spectrum, the spectrum was segmented into multiple scales. Each segment was fitted using a pseudo-Voigt function, with the absorption edge position. The GANDALF algorithm, a table-based deep learning approach, was employed to model the coordination environment of absorbing elements. The proposed method was validated using a previously published open-access dataset. For vanadium-containing samples, the model achieved R² values of 0.7837 on test sets with non-integer coordination numbers, whereas the random forest model only achieved 0.6328. Furthermore, our results highlight the significant importance of the post-edge peak when predicting coordination numbers using the full spectrum. Full article

The Internet of Things (IoT) has established an exceptional ecosystem of interconnected devices where a vast multitude of heterogeneous devices can communicate, collect, and share data for enhanced decision-making processes. To effectively analyze this immense volume of data, researchers have deployed advanced machine learning algorithms and deep neural networks. However, these approaches typically rely on centralized data storage for training, which raises significant privacy concerns. Federated Learning (FL) addresses this issue by allowing devices to train local models on their own data and share only model updates. Despite this advantage, FL remains vulnerable to several security threats, including model poisoning, data manipulation, and Byzantine attacks. Therefore, robust and scalable authentication mechanisms are essential to ensure secure participation in FL environments. This study provides a comprehensive survey of authentication in FL. We examine the authentication process, discuss the associated key challenges, and analyze architectural considerations relevant to securing FL deployments. Existing authentication schemes are reviewed and evaluated in terms of their effectiveness, limitations, and practicality. To provide deeper insight, we classify these schemes along two dimensions as follows: their underlying enabling technologies, such as blockchain, cryptography, and AI-based methods, and the system contexts in which FL operates. Furthermore, we analyze the datasets and experimental environments used in current research, identify open research challenges, and highlight future research directions. To the best of our knowledge, this study presents the first structured and comprehensive analysis of authentication mechanisms in FL, offering a foundational reference for advancing secure and trustworthy federated learning systems. Full article

A receipt information extraction task requires both textual and spatial analyses. Early receipt analysis systems primarily relied on template matching to extract data from spatially structured documents. However, these methods lack generalizability across various document layouts and require defining the specific spatial characteristics of unseen document sources. The advent of convolutional and recurrent neural networks has led to models that generalize better over unseen document layouts, and more recently, multi-modal transformer-based models, which consider a combination of text, visual, and layout inputs, have led to an even more significant boost in document-understanding capabilities. This work focuses on the joint use of a neural multi-modal transformer and a rule-based model and studies whether this combination achieves higher performance levels than the transformer on its own. A comprehensively annotated dataset, comprising real-world and synthetic receipts, was specifically developed for this study. The open source optical character recognition model DocTR was used to textually scan receipts and, together with an image, provided input to the classifier model. The open-source pre-trained LayoutLMv3 transformer-based model was augmented with a classifier model head, which was trained for classifying textual data into 12 predefined labels, such as date, price, and shop name. The methods implemented in the rule-based model were manually designed and consisted of four types: pattern-matching rules based on regular expressions and logic, database search-based methods for named entities, spatial pattern discovery guided by statistical metrics, and error correcting mechanisms based on confidence scores and local distance metrics. Following hyperparameter tuning of the classifier head and the integration of a rule-based model, the system achieved an overall F1 score of 0.98 in classifying textual data, including line items, from receipts. Full article