Search Results (443)

Cardiovascular diseases remain one of the most critical medical conditions worldwide, often resulting in severe complications that demand early and precise diagnosis. While previous studies have predominantly addressed cardiovascular disease prediction using single-label classification models, such approaches are insufficient to capture the multi-label and interdependent nature of myocardial complications. Motivated by these limitations, this research proposes a novel framework that leverages a Hybrid Shallow Neural Network (HSNN) architecture, designed to balance model complexity and computational efficiency. Additionally, a new feature selection algorithm, termed Multi-label Gini Importance (MGI), is introduced, which thoroughly evaluates and selects the most relevant features across all labels by employing a Gini-impurity-based mechanism. To further enhance the integrity of the dataset, a K-nearest-neighbors-based imputation strategy is utilized for addressing missing values. Experimental evaluations show the superiority of the proposed methodology, demonstrating significant advancements over traditional machine learning algorithms and multi-label classification techniques. The proposed framework achieves an outstanding F1-score of $91.2\%$ and a Hamming loss of $0.052$. To further validate the practicality of the proposed approach, the HSNN model was implemented on an Altera Arria 10 GX FPGA platform, demonstrating its hardware efficiency and real-time processing capability. The hardware implementation achieves a maximum operating frequency of 210 MHz, a low inference latency of 1.85 $\mathrm{\mu }$s, and a high throughput of 540,000 inferences per second, while consuming only 4.2 W of power. Additionally, the design maintains low resource utilization across logic elements, DSP blocks, and memory, confirming its scalability and efficiency for embedded deployment. These findings affirm the robustness, interpretability, and predictive efficiency of the proposed system, offering a substantial contribution toward the development of intelligent clinical decision-support mechanisms for cardiovascular disease diagnosis and prognosis. Full article

Cultural heritage (CH) sites are continuously exposed to pressures that can lead to deterioration. Continuous monitoring and anomaly detection (AD) enable early damage detection for preventive conservation. We propose a late-fusion AD framework combining NST-Net, a nested spatiotemporal neural network, with a semantic module encoding expert conservation rules, and evaluate it on five real-world CH monitoring datasets. NST-Net outperforms six state-of-the-art baselines on most sites, achieving, on average, 42% higher Average Precision than the best baseline per site under a synthetic evaluation protocol emphasizing sharp, short-duration anomalies. Dataset length serves as an important performance factor: the combined architecture and preprocessing pipeline particularly benefit longer campaigns, whereas they add little value on shorter ones. NST-Net also requires approximately 13 times less peak memory than baseline detectors, at a higher but still millisecond-scale latency. Further analysis shows that performance depends strongly on anomaly density, type, and severity. The semantic module passes label-free sanity checks and complements NST-Net effectively (complementarity index > 0.87), identifying conservation-relevant events missed by the deep model. These findings support late-fusion statistical–semantic frameworks for AD in CH monitoring. Full article

Random Forest and other tree-ensemble classifiers achieve high accuracy in network intrusion detection; however, their aggregate decision logic prevents analysts from auditing or deploying individual predictions as operational rules. Post hoc explanation methods introduce latencies incompatible with security operation center (SOC) requirements and produce conditions unsuitable for firewall configuration. Among the systems reviewed in this study, none unifies intrinsic explanation, rule deployment, ATT&CK attribution, cross-dataset validation, and adaptive feedback in one pipeline. This work presents a depth-limited Random Forest with deterministic, per-instance explanations at a fraction of gradient-based attribution latency. Complementary mechanisms generate analyst-deployable rule specifications, technique-level adversary attribution, and a feedback protocol that models label noise, missed reviews, and bounded correction budget. Evaluated on a large, multi-category network-traffic benchmark, the system attains high detection accuracy (macro recall $0.86$, driven substantially by the majority normal-traffic class at $68\%$ of flows) while sustaining throughput beyond SOC requirements; a stealthy reconnaissance-and-exploitation category remains markedly harder to detect under this class imbalance. Cross-dataset evaluation on a more recent benchmark attains strong performance after limited target-domain retraining. The adaptive feedback protocol yields a statistically significant false-positive reduction over repeated simulated reviews, requiring only modest weekly analyst effort. Together, these capabilities enable auditable and SOC-integrable detection pipelines. Full article

The Hydrogen Incidents and Accidents Database (HIAD) includes events initiated within hydrogen equipment, within a system that also contains hydrogen, or in a non-hydrogen system. This study characterized associations between the curator-assigned initiating system category and seven structured descriptors and multi-label root cause classes in HIAD 2.4. Analyses retained the original HIAD categories and used contingency-table tests, Cramér’s V, adjusted standardized residuals, Holm adjustment, and multinomial models. Of 1235 records with a known initiating system category, 825 (66.8%) were classified as hydrogen system-initiated, 99 (8.0%) as hydrogen-containing system-initiated, and 311 (25.2%) as non-hydrogen system-initiated. The largest associations were observed for initiating cause (V = 0.525) and supply chain stage (V = 0.492). Hydrogen-containing system records were most frequently classified in process-gas service (77.8%) and rupture with ignition (62.6%), whereas non-hydrogen system records more frequently involved unintended chemical hydrogen generation, impact/rollover/crash, no hydrogen release, and near misses. Material/manufacturing, installation, and job-factor labels were associated with lower adjusted odds of non-hydrogen-system versus hydrogen-system initiation. Associations generally persisted in higher-quality and post-2000 subsets. Because the compared fields were assigned by the same curators from the same narratives, the findings indicate within-database profile separation rather than the independent validation of incident mechanisms or accident rate estimation. Full article

Intelligent monitoring of diseases and insect pests in rice and wheat has evolved from handcrafted features and conventional machine learning to deep learning, multimodal data fusion, and time-series forecasting. This review compares data acquisition and representation, unimodal recognition, multimodal fusion, temporal prediction, and field generalization with respect to data requirements, task outputs, application contexts, and the strength of supporting evidence. Conventional machine learning remains valuable for small datasets, variable interpretation, and baseline comparisons, whereas deep learning extends monitoring from classification to detection, segmentation, pest counting, and severity estimation. Multimodal and temporal models further integrate phenotypic, physiological, environmental, and pest-monitoring information to predict future risk. However, many reported gains are weakened by inadequate spatiotemporal alignment, non-independent data partitioning, limited missing-modality tests, and insufficient cross-location and cross-year validation. Future research should prioritize standardized multisite, multiyear datasets; label-efficient, mechanistically informed, and trustworthy fusion methods; lightweight deployment; and prospective field trials that link model outputs to management decisions and production outcomes. Full article

Falls in older adults are a leading cause of injury, and the time spent on the floor afterwards is the stronger predictor of outcome. We present a low-cost system that detects the sustained lying posture following a fall. An FLIR Lepton 3.1R (160 × 120 px) and an int8-quantized FOMO detector run fully on an OpenMV RT1062 board; no image data leaves the node. Ten healthy adults, none in the training data, followed a structured posture protocol, yielding 3034 labelled frames. No fall, real or simulated, was recorded: lying is a proxy for a post-fall state, and claims are restricted accordingly. The detector produced an output on only 66.3% of labelled frames, with a strong class dependence (75.6% lying, 51.7% standing). End-to-end, binary lying-posture recognition reached a sensitivity of 0.723 (95% CI 0.637–0.812) and an F1 of 0.801. Every sustained lying bout of at least 20 s was flagged, but the deployed alert rule also produced roughly one hundred false alerts per hour of non-lying activity. Low-resolution thermal sensing is therefore a workable basis for long-lie detection; the limiting factors are the detector’s class-dependent miss rate and the alert logic, not the posture classifier. Full article

Accurate punctuation and capitalization are essential for the readability, interpretability, and structural coherence of machine-generated text. Their absence is particularly problematic in automatic speech recognition outputs and other forms of unstructured text, where missing punctuation and incorrect capitalization reduce both human readability and the effectiveness of downstream natural language processing tasks. This study proposes a hybrid XLM-RoBERTa–BiLSTM model for joint punctuation restoration and text capitalization on English-language data. The proposed architecture combines contextual representations produced by the multilingual pre-trained XLM-RoBERTa encoder with the sequential modeling capabilities of a bidirectional long short-term memory layer, followed by token-level classification in a unified label space. The model was trained and evaluated on a dataset derived from the IWSLT 2012 TED Talks corpus. Experimental evaluation on a held-out random test subset demonstrates strong performance. Excluding the dominant no-punctuation class, the model achieves an accuracy of 0.929, precision of 0.892, recall of 0.914, and an F1-score of 0.903. Including the dominant no-punctuation class increases these values to 0.961, 0.927, 0.919, and 0.923, respectively, reflecting the pronounced class imbalance in the dataset. Class-wise analysis shows high effectiveness for frequent punctuation classes and reliable capitalization prediction, whereas rare punctuation–capitalization categories remain more challenging because of their limited representation in the training and test subsets. Overall, the proposed hybrid XLM-RoBERTa–BiLSTM architecture achieves strong performance in joint punctuation restoration and text capitalization and represents an effective approach to improving the readability and structural quality of automatic speech recognition transcripts and other machine-generated text. Full article

Internet of Things (IoT) traffic anomaly detection is essential for limiting device compromise and large-scale attacks. Existing detectors may miss attack families absent from model development, while heterogeneous benign traffic makes it difficult to maintain a low false-positive rate (FPR). To address these two practical limitations, we propose the Mode-Calibrated Dual-Evidence Detector (MCDE). Its supervised branch estimates the probability that a sample is malicious from labeled benign and known-attack traffic, while its benign-deviation branch measures distance from multiple learned benign traffic modes, providing a complementary route for unseen attacks. MCDE maps the heterogeneous probability and distance scores to comparable empirical benign-tail evidence, normalizes each branch by its allocated share of the target FPR, and fuses them into an anomaly score. A disjoint held-out benign set determines the decision threshold. Equivalently, the fusion compares budget-adjusted benign-tail surprisal, linking the decision rule to empirical self-information. We further establish the conditions under which the budgeted fusion controls the nominal overall FPR. Family-hold-out experiments on IoT-23 and N-BaIoT validate MCDE. At a 1% target benign FPR, MCDE improves IoT-23 unseen recall over histogram-based gradient boosting from 85.77% to 90.85% and harmonic known–unseen recall from 91.82% to 95.07%, while maintaining a 0.96% benign-test FPR. It also achieves 99.87% unseen recall on N-BaIoT. Full article

This methodological diagnostic tests whether broad building descriptors support retrofit-class shortlists beyond popularity. We mapped 2623 New York City Local Law 87 audit rows to six classes with five decision states, trained on 2019–2022, validated on 2023, and tested on unseen 2024 property groups. Under natural recorded-label visibility, state-aware and equal-information global policies both achieved an observed-positive Recall@3 of 0.832. The state model returned the same Top-3 set for 99.7% of test rows; all evaluated natural-condition neural models missed every recorded insulation and window-upgrade positive. Light-emitting diode (LED) lighting and heating, ventilation, and air conditioning (HVAC) controls comprised 277 of 394 test positives, explaining why high recall did not demonstrate personalization. In a secondary synthetic 10% visibility stress test, excluding unlabelled entries from negative supervision improved recall over naive binary cross-entropy by 0.279; the natural state–naive difference was 0.002 (95% confidence interval (CI) [−0.007, 0.012]). Post hoc capacity, ontology, status-precedence, and stopping sensitivities did not establish reliable superiority over popularity. The benchmark diagnoses label misspecification and nearly constant shortlists. Full article

Aero-engine remaining useful life (RUL) prediction supports condition-based maintenance, yet most data-driven models assume fixed sensor availability. Power-supply, acquisition, or communication failures can invalidate this assumption. We propose the Remaining Useful Life Dual-Attention Robust Network (RUL-DARNet), which combines a convolutional neural network–long short-term memory (CNN–LSTM) backbone with training-stage whole-channel Sensor Dropout (SD), Mask-Aware (MA) feature attention, and temporal attention. SD exposes the model to reduced sensor sets, whereas MA excludes unavailable channels from feature-attention normalization using an explicit availability mask. Ten seeds and ten paired masks were evaluated across four Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) subsets under controlled synthetic sensor unavailability. Trajectory metrics use a 125-cycle label cap and equal engine weighting. At the prespecified FD001 40% Missing Completely at Random endpoint, RUL-DARNet attained an RMSE of 17.474 ± 1.410 cycles, compared with 19.023 ± 0.706 for SD-only. Adding MA after SD reduced RMSE by 1.549 cycles in nine of ten seeds after Holm correction. Benefits weakened or reversed under value-related missingness, multiple operating conditions, and several trajectory-level outages. Training-stage exposure accounts for most of the observed robustness, while mask-aware reweighting provides a smaller, conditional benefit within the tested C-MAPSS protocols when availability labels are reliable and the remaining channels retain degradation information. Full article

This study addresses a regime-conditioned question in electricity theft detection: when only monthly billing series and inspection-confirmed labels are available, which supervised model families recover irregular consumption without relying on the temporal resolution of advanced metering infrastructure (AMI)? The working sample comprises 4000 utility customers and 864 confirmed theft cases, each represented by 53 monthly kWh values from January 2021 to May 2025. After majority-class undersampling that retains all theft observations and construction of a balanced 1:1 learning set, eight classifiers are compared under an 80/20 stratified split: K-Nearest Neighbors, Decision Tree, Support Vector Machine, Random Forest, two dense multilayer perceptrons, Long Short-Term Memory, and a one-dimensional Convolutional Neural Network. Performance is assessed through threshold-optimized accuracy together with precision, recall, F1-score, the area under the receiver operating characteristic curve (AUC), and confusion matrices. On the hold-out test set, Random Forest and the compact dense network both reach an accuracy of 0.685; Random Forest attains the highest AUC (0.748) and F1-score (0.677). Even so, these models miss about one-third of the hold-out theft accounts (59 and 65 false negatives out of 173). Sequential deep models underperform on this short monthly regime. The results support ensembles and compact dense networks for monthly theft screening and indicate that AMI-oriented sequential gains do not transfer automatically to 53-point billing vectors under the present protocol. Full article

Automated alert triage can reduce Security Operations Center (SOC) workload, yet the validation-tuned thresholds deployed systems rely on carry no finite-sample control of their operational error rates and degrade unpredictably under distribution shift. We present a model-agnostic dual-threshold conformal deferral architecture: high-score alerts are auto-escalated under finite-sample marginal class-conditional control of the benign-escalation probability (budget α), low-score alerts are auto-closed under matching control of the threat-miss probability (budget β), and the rest are deferred to an analyst. It needs no retraining and closes an automatic zone rather than certifying what the calibration data cannot support. We evaluate it on a reinforcement-learning investigation agent in a simulated SOC and on four classifiers trained on CIC-IDS2017 and tested on CSE-CIC-IDS2018, using stratified 25,000-flow calibration and evaluation samples, with attack-type recall computed over the full 16.2-million-flow corpus. Pooling episodes from ten trained policies across two evaluation datasets, the architecture automated 73.7% of decisions at α = β = 0.01—a figure for that predefined pooled mixture rather than a per-policy or per-dataset guarantee—realizing benign auto-escalation and threat auto-close rates of 0.0099 and 0.0101 and deferring the hardest ~26% of alerts. After recalibration on labeled target-domain data, severe cross-dataset degradation appears not as a silent error but as sharply reduced certifiable automation, with deferral rising to 79–99% for the most affected classifiers. This visibility is a property of the recalibrated layer: thresholds left un-recalibrated after a shift continue to certify nothing while still deciding, so the architecture requires periodic recalibration on labelled target-domain alerts to deliver it. Substituting open-weight language models for the analyst inside the band failed a pre-specified criterion at every scale tested from 7B to 32B across two model families, with the discriminative signal flat in model size and far below the first-stage policy’s own. Full article

Automating the identification of International Classification of Diseases (ICD) codes from electronic health records (EHRs) presents a critical challenge, particularly for rare diseases where existing computational methods severely underperform due to extreme long-tail label distributions. To address this, we propose a multi-modal deep learning framework known as MMAC-Net, designed to enhance the retrospective assignment of ICD-9 codes to admissions involving rare pathologies. The model integrates unstructured clinical narratives with structured auxiliary data, specifically pharmacological prescriptions and microbiology events, using a convolutional attention-based architecture. Through a late fusion mechanism, it synthesizes attention-weighted textual representations with dense embeddings of the structured data types. Validation on the MIMIC-III dataset shows consistent improvements over a matched text-only baseline evaluated under an identical protocol. On the full dataset of 8930 ICD codes, the framework achieved a Micro-AUC of 0.997 and Precision@8 of 0.875. On the subset of admissions carrying at least 1 of 568 rare codes, adding the two structured modalities to the text encoder raises Macro-F1 from 0.011 to 0.084 and Micro-F1 from 0.368 to 0.513 relative to the text-only baseline, corresponding to relative increases of 6.69 and 0.39, respectively, while Precision@8 rises from 0.092 to 0.159 and Micro-AUC from 0.966 to 0.985. While extreme class imbalance remains a formidable obstacle, these findings underscore that incorporating structured clinical context partially mitigates the limitations of purely natural language processing approaches. Practically, the framework is intended as a decision-support component that presents a ranked shortlist of candidate codes to a human coder or clinician; by recovering rare codes that text-only systems miss, it targets the under-coding of low-prevalence conditions that degrades registry completeness and downstream epidemiological estimates. Full article

Driving style, reflected in recurrent patterns of acceleration, braking, speed selection, following distance, gear use, and lane-changing behavior, plays an important role in road safety and is also associated with fuel consumption, emissions, passenger comfort, and vehicle wear. This paper presents a structured review of recent research on driving style analysis, with particular emphasis on its relationship with road-safety outcomes and risk indicators. Following a PRISMA-oriented methodology, studies published between 2015 and 2025 were identified, screened, and synthesized to examine how driving styles are defined, detected, classified, and evaluated. The review shows a clear shift toward data-driven approaches, including feature-based machine learning and representation-learning methods using support vector machines, ensemble models, convolutional neural networks, recurrent neural networks, and hybrid deep learning architectures. Common data sources include smartphone inertial and GNSS signals, CAN/OBD vehicle data, telematics platforms, naturalistic driving datasets, and camera-based perception systems. Safety impact is most often assessed through crashes, near-miss events, traffic conflicts, time-to-collision measures, harsh maneuvers, and composite risk scores. Across the reviewed literature, aggressive and unstable driving patterns are generally associated with reduced safety margins and increased risk, although comparability remains limited by inconsistent label definitions, heterogeneous datasets, indirect safety proxies, and varied validation protocols. The paper also proposes a reproducible database architecture linking drivers, trips, driving events, and safety events to support transparent analysis, benchmark development, and future implementation in fleet monitoring, driver feedback, and connected vehicle applications. Full article

Automated sentiment analysis of online-education reviews is useful for understanding learner feedback. Classification-based methods usually capture only document-level polarity. They may miss aspect-level signals and may collapse to the majority class under the heavy imbalance typical of course reviews. When the task is reformulated as generation, document-level and aspect-level outputs can be unified. However, out-of-vocabulary aspect labels, parsing failures, and weakly grounded links between granularities may also be introduced. Multi-perspective and Holistic Evaluation T5 (MHE-T5), a model built on the Text-to-Text Transfer Transformer (T5), is proposed as a constrained multi-granularity generative model. It emits aspect-level and document-level sentiment in one schema. The model combines grammar/finite-state machine (FSM)-constrained decoding, a document–aspect consistency coupling with a proved alignment property, and a cross-granularity contrastive objective. The decoding guarantee is limited to schema parse-validity and closed-vocabulary conformity; it does not guarantee semantic correctness of the selected aspect or polarity. Across four datasets, including a large rating-derived Coursera corpus, two human-annotated education aspect-based sentiment analysis (ABSA) datasets, and the standard Multi-Aspect Multi-Sentiment (MAMS) benchmark, generative models improve macro-averaged F1-score (Macro-F1) over Bidirectional Encoder Representations from Transformers (BERT) by 0.36 to 0.61 on the three datasets that carry discriminative baselines. MHE-T5 attains the highest document-level Macro-F1 among the evaluated benchmarks while providing formal schema-level guarantees on the closed-vocabulary settings. A controlled comparison with DeepSeek-V3 on identical examples, used as a large language model (LLM) baseline, shows that the fine-tuned 220M model is a competitive schema-constrained fine-grained aspect extractor under the fixed protocol. Full article