Search Results (5,547)

Implantable Cardioverter-Defibrillators (ICDs) are life-sustaining devices used in the prevention of sudden death in patients suffering from advanced cardiac diseases. Although improvements in ICD technology and monitoring capabilities have been made, existing techniques are still not effective in predicting accelerated battery drain, thereby causing unplanned generator replacement and clinically significant device-related events. In this study, machine learning techniques were employed in the assessment of the early detection of ICD battery depletion risk using the collected device interrogation reports. The dataset used consisted of 32 devices in the training set and nine in the testing set. In order to mitigate the problem of a small sample size, a GMM-based data augmentation technique was applied solely to the training data, and actual devices were used for the testing data. Five supervised models, including Logistic Regression, Random Forest, SVM, CatBoost, and a Neural Network (MLP), have been utilized using a repeated stratified cross-validation and a separate held-out test data set. All the models have been tested for their performance using classification metrics. All models demonstrated variable performance with wide confidence intervals due to limited sample size. Support vector machines showed the highest cross-validation discrimination 0.889 ± 0.203, though uncertainty remains substantial given the small datasets (n = 41). From the feature analysis, it was found that atrial sensing amplitude, RV/LV capture threshold, output settings, and implant duration were the most important features for the prediction of high battery depletion risk. These findings suggest that changes in device parameters and implant age are associated with elevated battery depletion risk, supporting the feasibility of telemetry-driven risk stratification for device management. Full article

Interlayer bond strength is critical for ensuring the safety and durability of 3D-printed concrete (3DPC) structures. However, there remains a lack of real-time prediction methods addressing interlayer performance under the combined effects of interval time and environmental factors during the in situ printing process. To address this issue, this study conducted experiments considering various printing interval times and environmental conditions, incorporating monitoring of dielectric constant and water evaporation, alongside interlayer splitting tensile tests. By integrating the SHAP interpretability algorithm with nonlinear regression analysis, the results indicate that the printing interval time is the dominant factor inducing interlayer strength decay (with a contribution rate of 68.6%), while relative humidity emerges as the primary environmental variable (with a contribution rate of 21.3%). Mechanism analysis reveals that prolonged printing intervals intensify the hydration of the lower deposited layer, leading to reduced interfacial moisture content and loss of plasticity. Furthermore, environmental evaporation significantly regulates this process, with high-humidity environments notably mitigating the moisture loss and strength reduction caused by time delays. Based on the correlation mechanism between moisture and strength, a dimensionless general prediction model for 3DPC interlayer strength was established, incorporating printing interval time and an evaporation index (goodness of fit, R² = 0.96). Consequently, a digital twin quality inversion scheme based on companion specimen monitoring and printing timestamps was proposed. This study quantifies the intrinsic relationships among printing interval time, environmental conditions, and interlayer strength, offering a novel approach for determining the construction window and achieving non-destructive quality prediction for 3DPC in complex environments. Full article

Background: Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection and remains a leading cause of mortality in intensive care units (ICUs). Although the Sequential Organ Failure Assessment (SOFA) score is widely used for prognostic stratification, organ dysfunction represents a downstream manifestation of sepsis, whereas immune and inflammatory dysregulation may precede overt organ failure. Monocyte distribution width (MDW) is a novel hematological parameter reflecting monocyte activation and is approved for the diagnosis of sepsis; however, its prognostic value and potential role within composite biomarker models in critically ill surgical patients with sepsis remain incompletely defined. Methods: We conducted a prospective, observational, single-center pilot study in two surgical intensive care units between November 2022 and December 2023. Adult patients with sepsis defined according to Sepsis-3 criteria were enrolled. Laboratory and clinical variables—including MDW, neutrophil-to-lymphocyte ratio (NLR), C-reactive protein (CRP), procalcitonin (PCT), and SOFA score—were measured on admission and during the first five days of ICU stay. Patient-level median values across five days were used for analysis. The primary outcome was in-hospital mortality. Prognostic performance was assessed using receiver operating characteristic (ROC) curve analysis and logistic regression. A composite bioscore was constructed by combining dichotomized MDW, NLR, CRP, and PCT values. Results: Sixty patients were included; 24 (40%) died during hospitalization. Non-survivors were older and had significantly higher SOFA scores. MDW, NLR, CRP, and PCT were significantly higher in non-survivors. SOFA demonstrated the strongest discriminative ability for mortality prediction (AUC 0.839, 95% CI 0.730–0.948). Among biomarkers, NLR (AUC 0.741) and PCT (AUC 0.714) showed good discriminative performance, while MDW (AUC 0.690) and CRP (AUC 0.662) showed moderate discrimination; MDW exhibited the highest specificity (80.6%). In multivariable analysis with individual biomarkers, only SOFA remained an independent predictor of mortality. The composite bioscore demonstrated good discriminative ability (AUC 0.805) and, when evaluated alongside SOFA, remained independently associated with fatal outcome (OR 11.92, 95% CI 1.76–80.75); however, given the modest sample size and wide confidence intervals, this finding should be interpreted with caution. Repeated-measures correlation analysis revealed no strong collinearity among biomarkers. Conclusions: A composite bioscore incorporating MDW, NLR, CRP, and PCT provides prognostic information comparable to SOFA and remains independently associated with mortality. This approach may complement organ dysfunction-based assessment and support early risk stratification in sepsis. Full article

Background/Objectives: Vitamin B12 deficiency is a common but often underdiagnosed complication in patients with type 2 diabetes (T2D) undergoing long-term metformin therapy. Accurate early prediction could enable targeted screening and timely intervention. This study aimed to develop and interpret a machine learning model for predicting vitamin B12 deficiency in metformin-treated patients with T2D, using eXtreme Gradient Boosting (XGBoost). Methods: A retrospective cross-sectional study was conducted at a single endocrinology centre (La Rabta University Hospital, Tunis, Tunisia). Patients with T2D treated with metformin for at least three years were included (n = 257); those with conditions independently affecting vitamin B12 metabolism were excluded. Vitamin B12 deficiency was defined as a serum B12 level below 150 pmol/L or a borderline level (150–221 pmol/L) with concurrent hyperhomocysteinemia (>15 μmol/L). XGBoost was selected after comparison with Logistic Regression (L2), Random Forest, and Support Vector Machine on the same 5-fold stratified cross-validated pipeline. Hyperparameters were optimized via Bayesian search (100 iterations × 5-fold stratified cross-validation), with the Matthews correlation coefficient (MCC) as the primary optimization metric to account for class imbalance. Model interpretability was achieved using SHapley Additive exPlanations (SHAP). Discrimination and calibration were assessed on an independent test set using bootstrap 95% confidence intervals (2000 resamples). Results: Of 257 patients, 95 (37.0%) presented with vitamin B12 deficiency. On the independent test set (n = 52), the optimized XGBoost model achieved an ROC-AUC of 0.671 [95% CI: 0.514–0.818], sensitivity of 0.737 [95% CI: 0.533–0.938], specificity of 0.545 [95% CI: 0.375–0.710], MCC of 0.273 [95% CI: 0.018–0.517], and a Brier Score of 0.259. SHAP analysis identified HbA1c, microalbuminuria, autonomic neuropathy, BMI, DN4 score, and fasting glucose as the most influential predictors. Nonlinear SHAP interaction plots revealed an increased predicted risk in patients with low HbA1c combined with a high cumulative metformin dose. Conclusions: The XGBoost–SHAP framework provided interpretable predictions of vitamin B12 deficiency in patients with T2D on metformin, identifying key clinical profiles for targeted screening. External multi-centre validation is required before clinical deployment. Full article

This study establishes a facies-based framework for characterizing reservoir quality in the Upper Pannonian geothermal reservoirs of the Szentes field (Hungary). To evaluate vertical heterogeneity and optimize the selection of geothermal reinjection zones, an integrated core–log–statistical workflow was applied to data from boreholes SZT-1 and SZSZT-IX. The methodology combined petrophysical measurements, petrographic observations, and multivariate statistical analyses, including Hierarchical Cluster Analysis (HCA) and Linear Discriminant Analysis (LDA). The siliciclastic succession was classified into four distinct facies clusters representing a continuum of depositional energy regimes: Rolling, Graded Suspension with Rolling, fine-grained Suspension, and Uniform Suspension. The results demonstrate a dual control on reservoir quality: the primary pore framework is determined by depositional grain-size architecture and sediment transport processes, while mechanical compaction and diagenetic alteration subsequently modify pore connectivity and flow efficiency. Among the identified facies, deposits formed from Graded Suspension with Rolling represent the most favorable reservoir units, combining high porosity (up to 33%) with exceptionally high permeability (>1500 mD). In contrast, suspension-dominated facies deposited from Graded and Uniform Suspension exhibit significantly reduced permeability due to higher matrix content, cementation, and compaction. The results demonstrate that reservoir performance in the Szentes geothermal system is primarily controlled by facies-scale heterogeneity rather than by depth-based stratigraphic divisions alone. This integrated facies-based approach provides a predictive framework for extrapolating reservoir properties to uncored intervals and offers practical guidance for optimizing reinjection strategies and sustainable geothermal reservoir management. Full article

Tree ensembles—Random Forests (RFs) and Gradient Boosting Machines (GBMs)—often stabilize before all trees are evaluated. We study early stopping as a nonparametric change-point problem on prediction increments. The P²-STOP method family monitors a robust interquartile-range (IQR) scale of prediction increments online and stops when a relative-scale criterion is met. The default variant uses a rolling-window exact-quantile estimator ($O\left(w\right)$ memory), which provides a clean finite-sample stopping guarantee; a full-prefix P² streaming approximation ($O\left(1\right)$ memory) is available as a memory-light alternative. The stopping rule applies to both RFs and GBMs without model-specific distributional assumptions. On four RF benchmarks (MNIST, Covertype, HIGGS, and Credit Card Fraud), P²-STOP achieves 44.8% mean work reduction (range: 0.7–71.7%) with an accuracy change from $-0.53$ to $+0.02$ percentage points versus full-ensemble inference. On XGBoost ($T=500$), work reduction is dataset-dependent (41.4% on Covertype up to 89.0% on Credit Card), with corresponding accuracy trade-offs. Under random-tree contamination conditions (5%, 15%, and 25%), performance remains stable, whereas IQR-versus-standard-deviation baseline differences are mixed rather than uniformly dominant. Designed for compiled inference engines (e.g., C++/Numba), P²-STOP translates theoretical work reduction into consistent wall-clock speedups ($4.14\times$–$4.82\times$ versus compiled full RF on MNIST/Covertype/HIGGS for $T=500$). Native Python implementations serve purely as logical baselines due to loop overhead, while Credit Card exhibits the expected slowdown when work reduction is near zero. All comparisons use five seeds with 95% confidence intervals and seed-level paired tests. With only five seeds, inferential power is limited, and p-values should be interpreted cautiously. Relative to the Dirichlet RF baseline, our contribution is not larger RF-specific work reduction; it is a robust nonparametric IQR-scale stopping criterion, cast as a change-point/sequential-inference problem, that works as a post hoc wrapper across RF and GBM settings. Full article

The removal efficiency of wet scrubbers is governed by complex nonlinear interactions among operating parameters such as liquid level, airflow velocity, and dust concentration, making accurate real-time prediction challenging, which in turn leads to operational instability, increased energy consumption, and excessive emissions. To address this bottleneck, we first introduce multi-gene genetic programming (MGGP) to develop interpretable models quantifying multi-parameter coupling and predicting removal efficiency for PM₁, PM_2.5, PM₁₀, and TSP. Key input variables, including liquid level height, inlet airflow velocity, system pressure, and inlet dust concentration, were identified via correlation analysis. Explicit mathematical models were derived. Global sensitivity analysis using the elementary effect test (EET) identified inlet airflow velocity as most influential. Uncertainty quantification via quantile regression (QR) confirmed the model’s reliability with narrow prediction intervals and high coverage probabilities. MGGP offers a favorable balance of accuracy, generalization, and interpretability compared to extreme gradient boosting (XGBoost) and multiple nonlinear regression (MNR). Its explicit form quantifies parameter interactions, enabling efficient on-site monitoring with low computational cost. This study provides an interpretable prediction tool for intelligent wet scrubber operation, supporting cleaner production and refined control in complex industrial processes. Full article

Background: Adrenocortical carcinoma (ACC) is a rare and aggressive endocrine malignancy, for which population-level evidence regarding prognostic factors and survival conditions is limited. The available data mostly represent single-institution series, limiting their applicability. This study, therefore, assesses clinicopathological features and determines independent predictive variables of overall survival (OS) in patients with ACC using a population-based cohort. Methods: This retrospective observational cohort study used data from the Surveillance, Epidemiology, and End Results (SEER) Program between 2000 and 2022, initially identifying 1176 patients with ACC. Adult patients (≥18 years) with histologically confirmed ACC were identified using ICD-O-3 histology code 8370/3 and primary site code C74.0. Cases with zero-month survival, missing survival data, or identified only through autopsy or death certificate were excluded. To ensure dataset harmonization, patients with missing or indeterminate tumor grade and unknown stage were also excluded. After applying these inclusion and exclusion criteria, the final analytic cohort consisted of 267 patients. Data on demographic factors, stage of the disease, and treatment (surgery, chemotherapy, radiotherapy) were extracted. OS was evaluated using the Kaplan–Meier method, and independent prognostic factors were identified using Cox proportional hazards regression analysis. Results: The median OS was 54 months [95% confidence intervals (CI): 36–85]. The estimated 1-, 3-, and 5-year OS rates were 77%, 57%, and 48%, respectively. Survival differed significantly according to tumor grade, stage, and surgical treatment. In multivariable Cox regression analysis, increasing age [Hazard ratio (HR): 1.03, 95% CI: 1.02–1.04; p < 0.001], high tumor grade (HR: 2.21, 95% CI: 1.43–3.41; p < 0.001), and distant-stage disease (HR: 3.24, 95% CI: 1.95–5.38; p < 0.001) were independently associated with an increased risk of mortality, whereas surgical treatment was associated with improved survival (HR 0.53, 95% CI 0.30–0.93; p = 0.028). Chemotherapy and radiotherapy were not significantly associated with mortality. Conclusion: In this SEER-based cohort of patients with adrenocortical carcinoma, older age, high tumor grade, and distant-stage disease were independently associated with worse OS, whereas documented receipt of surgery was associated with longer OS. Treatment-related associations should be interpreted cautiously in view of the inherent limitations of registry-based data. Further prospective multicenter studies are needed to confirm these findings. Full article

We propose a fully nonparametric empirical autoregressive copula framework for univariate time series, designed to capture nonlinear and asymmetric serial dependence while exactly preserving the empirical marginal distribution. The method decouples marginal behavior from temporal dependence by (i) constructing a shape-preserving empirical marginal via monotone interpolation and mapping observations to the unit interval, and (ii) estimating the lag–lead dependence through a nonparametric conditional AR(1) copula density on ${(0,1)}^2$. To ensure stable estimation near the boundaries, we employ reflection-based kernel methods that mitigate edge effects and yield well-behaved conditional densities on the unit support. Forecasts are obtained from the implied conditional predictive density: we compute point forecasts either as conditional modes (maximum a posteriori) on the copula scale or as conditional means, and then back-transform exactly using the empirical quantile function, guaranteeing marginal fidelity and support-respecting predictions. Empirically, we evaluate the approach on three CBOE volatility indices (VIX, VXD, and RVX) and benchmark it against linear ARMA models, copula-based parametric competitors, and state-space/heteroskedasticity baselines (Local level, TVP–AR, and ARMA–GARCH). The results highlight that modeling the full conditional transition density nonparametrically can deliver competitive—often best or near-best—forecast accuracy across horizons, particularly in the presence of pronounced volatility regimes and asymmetric adjustments. Full article

Under dynamic loading conditions, the output voltage of proton exchange membrane fuel cells (PEMFCs) exhibits nonlinear degradation characterized by non-Gaussian fluctuations, abrupt changes, and long-range temporal dependence, which are difficult to model using conventional short-correlation or remaining useful life (RUL) prediction approaches. To capture both historical dependency and stochastic jump behavior, this study proposes a SHAP-driven mechanism–data fusion fractional stochastic degradation model based on fractional Brownian motion (fBm) and fractional Poisson process (fPp) for degradation trend forecasting. A terminal voltage mechanism model considering activation, ohmic, and concentration polarization losses is first established, and SHapley Additive exPlanations (SHAP) analysis is employed to quantify the contributions of multi-source operational variables and enhance interpretability. The Hurst exponent is then used to verify long-range dependence and jump characteristics in the voltage sequence. Subsequently, fBm is integrated with a fPp to construct a unified stochastic degradation framework capable of jointly describing continuous decay and discrete abrupt variations, enabling multi-step probabilistic prediction with confidence intervals. Validation on the publicly available FCLAB FC1 and FC2 datasets shows that the proposed model achieves superior overall performance under both steady and dynamic conditions, with MAPE/RMSE/R² of 0.027%/0.00178/0.9895 and 0.056%/0.00259/0.9896, respectively, outperforming fBm, Wiener, WTD-RS-LSTM, and CNN-LSTM methods. The proposed approach provides accurate and interpretable degradation forecasting for PEMFC health management and maintenance decision support. Full article

Hypertension is a leading global health risk and requires accurate and continuous monitoring for effective management. Although photoplethysmography (PPG) is a promising non-invasive modality for cuffless blood pressure (BP) assessment, many existing approaches (especially raw-signal deep learning) are vulnerable to data leakage, overfitting on small datasets, limited interpretability, and poor performance on minority BP stages. To address these limitations, we propose a robust and physiologically grounded framework for multi-class BP stage classification based on interpretable PPG features. Our approach centers on a comprehensive multi-domain feature engineering pipeline that extracts 124 PPG features, including demographic, morphological, functional decomposition, spectral, nonlinear dynamics, and clinical composite indices. We apply rigorous preprocessing and feature selection prior to model training. We validate the framework on two datasets: PPG-BP dataset (657 segments, 4 classes) for benchmarking and PulseDB (283,773 segments, 3 classes) to assess scalability. We evaluate the proposed framework using a segment-level train/test split, appropriate for assessing intra-subject BP tracking after initial personalization. For the PulseDB dataset, this follows the protocol established by the dataset creators, while for the PPG-BP dataset, it enables direct comparison with prior work given practical dataset constraints. On PPG-BP, LightGBM trained on the selected features achieved macro-F1 = 0.78 and accuracy = 0.74, outperforming comparable deep-learning models. On the PulseDB, a custom Residual MLP achieved accuracy = 0.81 and macro-F1 = 0.79, supporting generalization at scale. These results show that the proposed feature-based approach can outperform complex end-to-end deep-learning models on small datasets while providing improved interpretability. This work establishes a reliable and transparent pathway toward clinically viable continuous BP staging, moving beyond black-box models toward physiologically grounded decision support. Ablation analysis reveals that engineered features provide most of the predictive power (F1 = 0.911), while raw PPG features alone achieve modest performance (F1 = 0.384). For the minority hypertension stage 2 (HT-2) class, a bootstrap 95% confidence interval of [0.762, 1.000] is reported, reflecting uncertainty due to limited sample size. Full article

Background/Objectives: Peak cough flow (PCF) is an objective measure of cough effectiveness after stroke, but biomarkers reflecting physiological vulnerability related to reduced PCF are not well established. We investigated whether bone turnover markers (BTMs)—C-terminal telopeptide of type I collagen (CTX) and procollagen type 1 N-terminal propeptide (P1NP)—were associated with PCF in subacute ischemic stroke. Methods: In this retrospective study, 112 patients admitted within 21 days of stroke onset had fasting morning CTX and P1NP measured by electrochemiluminescence immunoassay, and PCF measured within 72 h of admission. Associations were assessed using Spearman correlation and multivariable linear regression with BTMs standardized (per 1 standard deviation increase), adjusting for age, sex, body mass index, onset-to-admission days, National Institutes of Health Stroke Scale score, Korean version of the Modified Barthel Index, estimated glomerular filtration rate, smoking status, and brainstem lesion. Results: CTX showed an inverse correlation with PCF (rho = −0.469; p < 0.001) and remained independently associated with lower PCF after multivariable adjustment (β = −42.32 L/min; 95% confidence interval, −56.12 to −28.52; p < 0.001), whereas P1NP showed weaker associations. In secondary outcome analyses, higher CTX was associated with low PCF (PCF < 160 L/min), aspiration pneumonia, and longer length of stay. Conclusions: Higher CTX levels were independently associated with lower peak cough flow and selected respiratory-related outcomes in this retrospective cohort. These findings are hypothesis-generating, do not imply prognostic validation, and warrant confirmation in prospective multicenter studies assessing incremental predictive value. Full article

Introduction: In adult patients undergoing deceased donor kidney transplantation, anesthesia management impacts graft function and survival and is influenced by various donor and recipient clinical factors. The aim of this study was to describe the perioperative factors and to evaluate their association with delayed graft function (DGF) during the first seven days after transplantation. Materials and Methods: This cross-sectional study of adult patients who underwent deceased donor kidney transplantation at a tertiary care hospital from 2022–2023 was performed to evaluate pre-, trans- and postoperative patient’s characteristics. Comparisons or association tests were implemented between patient characteristics grouped by the absence or presence of DGF. In the case of the variables with clinical relevance, univariate and multivariate logistic models were constructed to evaluate the predictive capacity of these variables to predict delayed graft function. Crude and adjusted odds ratio (ORs) with 95% confidence intervals were calculated for each variable. Results: DGF was present in 25/69 (36.23%) patients. The anesthesia time was significantly longer (310.28 vs. 273.55 min; p = 0.043) and the post-transplantation stay was significantly longer (11.04 vs. 8.11 days; p < 0.001) in patients with delayed graft function. In univariable analyses, male sex (p = 0.018), platelet count (p = 0.025), and surgical time (p = 0.062) showed significant or borderline associations with DGF. In the multivariable model, male sex remained independently associated with DGF (adjusted OR 10.64; 95% CI 1.23–92.1; p = 0.031). Platelet count (per 50 × 10³ µL increase) demonstrated a borderline inverse association (adjusted OR 0.57; 95% CI 0.32–1.02; p = 0.057). Conclusions: Our results suggest that male sex was independently associated with delayed graft function after deceased donor kidney transplantation, while platelet count showed a borderline association. Full article

Low-cost air quality sensors enable dense urban monitoring networks but require calibration against reference-grade instruments. While machine learning calibration is well-established for co-located sensor pairs, applying these calibrations to sensors deployed far from any reference station—the operational reality for most network sensors—lacks systematic methodology. We address this gap using 24 months of hourly data (August 2023–July 2025) from Sofia, Bulgaria, where five official reference stations (Executive Environmental Agency) operate alongside 22 AirThings low-cost sensors, four of which are co-located. Random Forest models achieved $R^2\in (0.53,0.75)$ across PM_2.5, PM₁₀, NO₂, and O₃, representing from 40% (for O₃) to 408% (for PM_2.5) improvement over Multiple Linear Regression baselines. Using leave-one-station-out spatial cross-validation, we derived pollutant-specific uncertainty growth rates ($\alpha$) from 3.84% to 5.62% per km, characterizing how calibration uncertainty increases with distance from reference stations (statistically significant for PM₁₀ and O₃, $p<0.05$). Applied to 18 non-co-located sensors, the framework generated 1.2 million calibrated hourly measurements with 95% prediction intervals over the study period. Co-location sites spaced 6 km apart achieve a less than 30% uncertainty increase at network midpoints, within EU Air Quality Directive thresholds for indicative monitoring. These empirically derived $\alpha$ parameters enable network planners to predict measurement reliability at arbitrary sensor locations without ground-truth validation, providing evidence-based guidance for cost-effective hybrid monitoring network design. Full article

Under high renewable energy penetration, wind power forecasting faces pronounced challenges due to strong randomness and uncertainty, making conventional point-forecast-centric paradigms insufficient for risk-aware and reliable power system scheduling. An uncertainty-aware temporal transformer framework for wind power forecasting is presented, integrating probabilistic modeling with deep temporal representation learning to jointly optimize prediction accuracy and uncertainty characterization. Crucially, rather than treating uncertainty quantification merely as a post-processing step, the central conceptual contribution lies in modularizing uncertainty directly within the attention mechanism. A probability-driven temporal attention mechanism is incorporated at the encoding stage to emphasize high-variability and high-risk time slices during feature aggregation, while a multi-quantile output and interval modeling strategy is adopted at the prediction stage to directly learn the conditional distribution of wind power, enabling simultaneous point and interval forecasts with statistical confidence. Extensive experiments on multiple public wind power datasets demonstrate that the proposed method consistently outperforms traditional statistical models, deep temporal models, and deterministic transformers, as validated by formal statistical significance testing. Specifically, the method achieves an MAE of $0.089$, an RMSE of $0.132$, and a MAPE of $10.84\%$ on the test set, corresponding to reductions of approximately $8\%$–$10\%$ relative to the deterministic transformer. In uncertainty evaluation, a PICP of $0.91$ is attained while compressing the MPIW to $0.221$ and reducing the CWC to $0.241$, indicating a favorable balance between coverage reliability and interval compactness. Compared with mainstream probabilistic forecasting methods, the model further reduces RMSE while maintaining coverage levels close to the 90% target, effectively mitigating excessive interval conservatism. Moreover, by adaptively generating heteroscedastic intervals that widen during high-volatility events and narrow under stable conditions, the model achieves a highly focused and effective capture of critical uncertainty information. Full article