Search Results (16,630)

Reliable and rapid forecasting of wind turbine output power is vital for operators, particularly day-ahead and intraday market scheduling and reserve allocation. However, the inherent unpredictability, intermittency, and volatility of wind turbine output make forecasting processes difficult. To address this challenge, this study proposes a Sinusoidal Representation Network (SIREN)-based forecasting model for high-accuracy, rapid direct multi-horizon forecasting of wind turbine output power. SIREN is selected due to the periodic and symmetrical mathematical structure of its sinusoidal activation function, which allows the model to represent both low-frequency trends and high-frequency sudden changes in wind energy data. To improve data quality, compensate for asymmetric fluctuations in wind data, and provide more suitable inputs for SIREN training. Several preprocessing steps are utilized before feeding the data into the model. The proposed preprocessing step includes a moving median filter, robust scaling based on median and interquartile range, Winsorizing clipping, and a Hampel filter to reduce the effects of instantaneous noise, outliers, and local peaks without disrupting temporal continuity. Subsequently, a Savitzky–Golay smoothing is applied to attenuate high-frequency measurement noise while preserving curvature, local peaks, and physically meaningful short-term dynamics in the data. The sliding-window approach is used to formulate the multi-horizon forecasting problem directly, and a direct h-step-ahead forecasting architecture is designed, preserving structural symmetry in the time series. The SIREN is trained and tested using MATLAB with the help of two different datasets: Dataset-1 has a 10 min resolution for 1 year, and Dataset-2 has a 1 h resolution for 15 years. The forecast horizon parameter h is considered separately for each step, and the proposed SIREN is independently trained, validated, and tested for each target horizon while maintaining chronological order. The results demonstrate that the proposed model is able to yield high forecast performance for a wide spectrum of horizons ranging from 10 min to 15 days. The accuracy of the proposed model for Dataset-1 is R² of 99.6%, MSE of 0.085%, MAE of 1.7%, and MAPE of 12%, while for Dataset-2, the accuracy is R² of 98.8%, MSE of 0.3%, MAE of 3.6%, and MAPE of 23%. Ablation and sensitivity analyses are conducted to evaluate the impact of the basic components used in the proposed model on forecasting performance. In addition, combative experiments are performed using traditional time series, ML, and DL forecasting techniques to better assess the contribution of the model. The obtained results show that the SIREN-based direct forecasting approach provides strong learning capability, as well as high forecasting accuracy, for both high-resolution and low-resolution wind power data. Overall, its ability to capture the symmetric and periodic characteristics inherent in wind turbine power data makes it a promising alternative for multi-horizon wind power forecasting applications. Full article

A recently developed high-strength nickel-based corrosion-resistant alloy has attracted increasing interest for drilling and production operations in unconventional oil and gas fields. Owing to its high resistance to media containing H₂S, CO₂ and chloride ions, together with its ultra-high strength and favorable strength–toughness balance, this alloy is suitable for demanding service conditions. During hot working, dynamic recrystallization (DRX) governs deformation softening, grain refinement and the subsequent microstructural state, and thus has a direct influence on final properties. In this work, isothermal compression experiments were conducted on this ultra-high-strength nickel-based corrosion-resistant alloy using a Gleeble thermal simulator at 1000–1150 °C and strain rates of 0.01–10 s⁻¹. Electron backscatter diffraction (EBSD) was used to quantify grain size, grain-boundary misorientation, kernel average misorientation (KAM) and the DRX volume fraction. The results indicate that higher deformation temperature generally accelerates DRX, lowers the KAM value and increases the recrystallized-grain fraction. Under a constant deformation temperature, the DRX volume fraction changes non-monotonically with strain rate, showing an initial increase followed by a decrease. Based on the EBSD-derived DRX fractions, linear and quadratic single-parameter models using the Zener–Hollomon parameter were examined first, but neither provided satisfactory fitting accuracy. A two-variable empirical model was therefore formulated for a fixed true strain of ε = 0.92 by considering deformation temperature and strain rate separately. The predicted values agree well with the experimental data, giving R² = 0.91278 and an average relative error of 8.53%. The proposed model captures the main variation tendency of the DRX volume fraction within the studied processing window and provides a useful basis for microstructure control and hot-working parameter design for ultra-high-strength nickel-based corrosion-resistant alloys. Full article

The increasing power density of modern electronics demands more efficient thermal management. Heat sinks with sinusoidal fins remain understudied, and the combined effect of perforations and variable fin spacing on transient performance has not been systematically quantified. This numerical study, conducted using ANSYS Fluent 2025 R2, analyzes three sinusoidal fin configurations under forced convection (3–5 m/s): solid fins (Case A), perforated fins (Case B), and perforated fins with alternating spacing of 2 mm and 4.5 mm (Case C). The base was maintained at 60 °C during a 20 s transient period. A mesh with an average skewness of less than 0.25 ensured numerical convergence. Case B showed remarkable uniformity in the base temperature (variations < 1 °C), in contrast to Case A (variations of up to 14.17 °C), due to a thermal boundary layer restart effect induced by the perforations. Case C reached the highest heat dissipation temperatures (up to 54.64 °C at 3 m/s), representing a 47.2% increase compared to Case A, indicating more effective heat extraction with this type of separate fin. The critical transient window occurs within the first 5 s (>85% of the total temperature rise). A vertical temperature gradient of 1.19 °C/mm was observed near the base. Although the perforations reduced the heat transfer area by 5.94%, the induced turbulence compensated for this loss. Sinusoidal fins with perforations and variable spacing significantly improve convective heat removal. Full article

Accurate, automated delineation of adult diffuse gliomas from multi-parametric magnetic resonance imaging (mpMRI) is central to quantitative neuro-oncology. Volumetric 3D networks dominate the BraTS leaderboard but require expensive GPUs, long training cycles, and provide diminishing returns relative to their compute budget. Slice-wise 2D models, by contrast, discard inter-slice context that is informative for thin tumor rims and small enhancing foci. We introduce GDNet, a 2.5D multimodal MRI segmentation framework for adult glioma evaluated on the BraTS 2024 cohort. GDNet consumes a stack of three adjacent axial slices from the four standard BraTS modalities (T1, T1ce, T2, FLAIR) as a 12-channel input to a compact U-shaped encoder–decoder with Group Normalization and predicts whole tumor (WT), tumor core (TC), and enhancing tumor (ET) masks for the central slice. The training pipeline pairs the 2.5D backbone with: (i) Exponential Moving Average (EMA) of model weights with decay 0.999, (ii) mixed tumor-aware slice sampling (p_tumor = 0.50), (iii) a compound Cross-Entropy + Soft-Dice loss, and (iv) AdamW with warm-up plus cosine annealing under Automatic Mixed Precision. We performed a systematic, step-by-step ablation covering a 2D baseline, EMA + mixed sampling, tumor-centered crop fine-tuning, a GDNet-inspired architectural integration, a region-aware loss, 3-slice and 5-slice 2.5D inputs, and connected-component post-processing, and we report multi-seed results to quantify reproducibility. On the held-out BraTS 2024 test partition, the final 3-slice 2.5D GDNet achieved positive-only Dice scores of 0.791 ± 0.000 (WT), 0.736 ± 0.003 (TC), 0.654 ± 0.004 (ET), and a mean foreground positive-only Dice of 0.820 ± 0.000 across seeds; the all-slice mean foreground Dice exceeded 0.927 ± 0.000. Validation positive-only scores were 0.805 ± 0.002 (WT), 0.757 ± 0.004 (TC), 0.683 ± 0.009 (ET). The inter-seed standard deviation was small for every region (≤0.01 Dice points), indicating low inter-seed variance across the two seeds evaluated; with only two seeds, we regard this as preliminary evidence of training stability rather than a strong reproducibility claim. The ablation isolated EMA + mixed tumor sampling and the 2.5D context window as the dominant sources of improvement; notably, a GDNet-style architectural integration with a region-aware loss did not outperform the simpler 2.5D U-Net on positive-only WT/TC/ET, and light post-processing improved only all-slice Dice. A failure-mode audit found that the residual catastrophic predictions are concentrated on a small minority of diffuse, infiltrative tumors with mass effect. Conclusions: Carefully engineered training strategies, tumor-aware sampling, EMA stabilization, and a modest 2.5D context window recover a substantial fraction of the accuracy of much heavier 3D networks at a fraction of the compute, are reproducible across seeds, and outperform a heavier GDNet-inspired architectural variant on the same data. GDNet is therefore a practical and, pending external validation, potentially clinically deployable framework for multimodal glioma segmentation on workstation-class GPU hardware. Full article

Grocery shopping remains a heavily car-dependent activity in urban areas, even for short-distance trips within residential neighbourhoods. A primary barrier to shifting toward active travel (walking or cycling) is the physical burden of carrying heavy or bulky goods. This study investigates whether a retailer-managed home delivery service could encourage consumers who currently rely on motorised modes for grocery shopping to shift towards active travel while preserving the in-store shopping experience. The analysis focuses on urban Italian consumers who currently use motorised modes for grocery shopping. Using a Stated Preference (SP) experiment and a Mixed Logit (MMNL) model (n = 88), we analyse the conditions under which such a service may encourage the adoption of active travel modes and support proximity-based shopping patterns. Given the exploratory nature of the study and the small, non-representative sample, the findings should be interpreted as preliminary evidence for urban motorised grocery shoppers rather than as representative of the Italian population. The results indicate a substantial willingness among respondents to adopt the proposed service configuration. Delivery time, service cost, and the availability of delivery time-window selection emerge as critical factors influencing consumers’ choices. Acceptance of the service is also influenced by perceptions of walking and cycling infrastructure quality, trust in the integrity of delivered groceries, preferences for local products, and concerns regarding the working conditions of delivery personnel. Additionally, the model reveals significant heterogeneity in preferences regarding delivery by drone/autonomous vehicle and a 100% reduction in greenhouse gas emissions relative to conventional motorised transport. Younger respondents exhibit a more favourable attitude towards automated delivery technologies, while differences in the valuation of environmental benefits emerge between male and female respondents. The findings suggest that retailer-managed home delivery may represent a promising mechanism for encouraging active travel among current motorised grocery shoppers, while maintaining consumers’ relationship with neighbourhood retail services. These results provide retailers and urban policymakers with valuable insights, suggesting that appropriately designed delivery services may support more sustainable and proximity-oriented shopping behaviours. Such services could potentially contribute to maintaining the accessibility and vitality of neighbourhood retail activities, particularly in ageing urban contexts. Full article

Soybean (Glycine max) germination is highly sensitive to neutral salt stress. Although sodium chloride (NaCl) and sodium sulfate (Na₂SO₄) co-exist in nature, their distinct phytotoxic mechanisms remain severely under-investigated. In this study, 50 germplasm accessions were systematically screened, identifying R014 as highly salt-tolerant and R120 as highly sensitive. Phenotypic and dynamic antioxidant monitoring (0–72 h) established 48 h as the critical tolerance window, revealing that Na₂SO₄ induces complex physical damage (crystallization) and osmotic injury, with its ionic toxicity significantly exceeding that induced by NaCl. Crucially, R014 effectively maintained peak activities of antioxidant enzymes (SOD, POD, CAT) to combat these specific stressors. By integrating deep RNA sequencing with weighted gene co-expression network analysis (WGCNA) using 48 h radicle data, significant transcriptomic reprogramming was revealed. WGCNA robustly isolated 35 functional modules, located five key phenotypic clusters, and defined three major hub genes (Glyma.11G101900, Glyma.17G185000, and Glyma.20G247850) that regulate calcium signaling. Verified by qRT-PCR, this study suggests the differential physiological and molecular architectural characteristics between chloride and sulfate toxicities, providing precisely targeted genetic loci for the breeding of salt-tolerant soybean. Full article

Wind turbine gearbox condition monitoring increasingly combines Supervisory Control and Data Acquisition (SCADA) data with Explainable Artificial Intelligence (XAI) for predictive maintenance. However, current XAI applications report attributions as static or globally aggregated feature-importance results. Such representations do not reveal when fault-related variables emerge, how dominance shifts between features, or how the explanatory structure evolves as degradation progresses. This limits their value for time-resolved diagnostic interpretation. To address this gap, this study proposes the Temporal Attribution Matrix (TAM), a temporal interpretability framework that tracks the evolution of XAI-derived feature importance across degradation periods. The central hypothesis is that temporal attribution patterns contain diagnostic information not captured by static feature-importance summaries. TAM was applied to a three-year SCADA dataset from Fuhrländer FL2500 wind turbines using XGBoost-SHAP and 1D-CNN Grad-CAM within sliding weekly windows. Four temporal measures were derived: feature onset time, dominance transition, attribution entropy, and cross-model consistency. Both XAI methods independently identified gearbox bearing temperatures 451 and 152 as the most influential features. TAM further revealed a synchronized thermal-feature onset on 23 October 2012, 14 SHAP dominance transitions compared with 70 Grad-CAM transitions, and a moderate cross-model Spearman correlation of 0.488. Secondary validation using WT82 confirmed TAM’s applicability beyond a single turbine. These results demonstrate that TAM extends static XAI by producing time-resolved degradation narratives for SCADA-based wind turbine predictive maintenance. Full article

False base station (FBS) attacks exploit the attach window before the network authenticates to the device. Existing User Equipment (UE)-side detectors typically need either labeled attack data, which is scarce and does not generalize to unseen attacks, or models too heavy for the resource budget of a smartphone or embedded endpoint. This study presents a lightweight UE-side detector built on a first-order Markov chain over a four-tuple state of packet type, direction, message identifier, and access-network type. A single counting pass fits the 119 KB chain, and thresholds are derived from normal traffic, so no attack labels are consulted. The capture path requires root and Qualcomm modem diagnostic access. Attacks surface as low-probability transitions, rare field values, and anomalous pacing, fused into a per session verdict with per-message attribution. On 192 commercial, testbed, and public LTE and 5G captures, the detector flags 51 of 53 attacks at an F1 of 88.70% in leakage-free leave-one-session-out evaluation and 96.23% once calibration covers the scored sessions. In five-fold cross-validation its F1 of 86.21% trails the strongest supervised baselines by margins that are not statistically significant, and it records the lowest latency (0.46 ms) and smallest working set (8.8 MB) among the eleven detectors benchmarked. Full article

Extreme precipitation and compound hazards can increase the risk of inundation and accidental release of untreated effluent from water treatment facilities, with potential downstream impacts within a short emergency-response window. Few studies have linked site-scale surface-runoff behavior, feasible emergency-response scenarios, and resilience-based decision support for critical water infrastructure. This study presents a GIS-based scenario-comparison framework that couples high-resolution surface-runoff simulation with an AHP-informed resilience interpretation to evaluate untreated effluent runoff and temporary flood-defense strategies at a water treatment plant in Jeollabuk-do, South Korea. A 1 m digital elevation model derived from drone-based LiDAR data was used in ArcGIS Pro to simulate two-dimensional unsteady surface-runoff propagation, producing water-depth and flow-velocity fields at 30 s intervals over 20 min. Three scenarios were compared under identical topographic, release, and hydraulic assumptions, no response, primary defense-line deployment, and secondary defense-line deployment, adding a 335 m barrier along the downstream road. Under the no-response scenario, released water reached the river after approximately 6 min, with a cumulative river inflow of 329.27 m³. The primary defense line reduced cumulative river inflow by 16.8%, and the secondary defense line by 78.2%, while delaying river arrival to 8 min and 30 s. An approximate surface-water balance and time-series analysis showed that the defense lines primarily redistribute water into temporary upstream storage rather than eliminate it. The simulation-derived indicators were linked to four resilience components whose relative importance was estimated using the Analytic Hierarchy Process (AHP) from 205 expert and practitioner responses, which identified recovery speed as the highest-priority component; the weighted normalized indicators are summarized as a transparent scenario-level composite resilience indicator that increases from the no-response to the primary and secondary defense-line scenarios. Because the stormwater drainage network, pollutant transport, and operational deployment uncertainties were not explicitly modeled, the results should be interpreted as a comparative assessment of water-volume transport risk rather than a deterministic prediction of inundation or pollution impact. Within these stated assumptions, the results indicate that a strategically placed secondary defense line can substantially reduce downstream river inflow and secure additional response time, providing preliminary decision support for disaster-risk reduction and emergency-response planning at critical water infrastructure. Full article

In this article, we investigate whether the crash of the German DAX index following the U.S. tariff announcement in April 2025 is consistent with pre-existing endogenous market fragility rather than a purely exogenous shock. The analysis is conducted within the Log-Periodic Power Law (LPPL) framework using the Filimonov–Sornette (FS) specification, complemented by shrinking-window estimation, Ornstein–Uhlenbeck residual diagnostics, surrogate time-series analysis, and a GARCH-based Monte Carlo false-positive assessment. The results reveal a statistically stable critical period preceding the observed market collapse, within which the tariff announcement occurred and was followed by a pronounced market decline. Overall, the findings suggest that the market operated in a regime of elevated systemic fragility, where the tariff announcement may have acted as a triggering event within an already critical state. This study contributes to the literature on financial critical phenomena by providing evidence that LPPL-based critical windows may be interpreted as periods of heightened systemic vulnerability rather than precise crash forecasts. From a risk management perspective, such periods may be informative for identifying conditions under which markets are particularly sensitive to external disturbances. Full article

Polyurethane-modified epoxy crack sealants can combine the cohesive strength of epoxy networks with the flexibility required for asphalt pavement crack repair. However, their selection under different winter pavement-temperature conditions requires an integrated evaluation of workability, low-temperature transition, dimensional stability, aging resistance, and interfacial adhesion. In this study, six ambient-curing polyurethane-modified epoxy crack sealants (EUPC) were prepared and assessed under representative winter pavement-temperature conditions, with SBS-modified asphalt used as a reference. All EUPC formulations satisfied the 30 min construction-window requirement, showed T_g values below 0 °C, T_5% values above 300 °C, and curing volume shrinkage no higher than 3.0%. After moisture–oxygen–ultraviolet coupled aging, the formulations retained a substantial proportion of both tensile strength and elongation, with EUPC-3/EUPC-4 showing a relatively balanced strength–ductility response. Compared with SBS-modified asphalt, the climate-matched EUPC formulations provided higher direct tensile adhesion, oblique shear adhesion, and flexural–tensile repair recovery. Overall, EUPC-1/EUPC-2, EUPC-3/EUPC-4, and EUPC-5/EUPC-6 are more suitable for mild, cold, and severe low-temperature winter conditions, respectively. Full article

Background: Early rehabilitation after pediatric ischemic stroke may support neuroplasticity and improve long-term functional outcomes. However, rehabilitation practices remain heterogeneous, and evidence-based recommendations regarding the optimal timing and intensity of intervention are limited. Objectives: This scoping review aimed to evaluate the available evidence regarding early rehabilitation after pediatric ischemic stroke, identify prognostic factors associated with functional recovery, summarize current therapeutic approaches, and highlight gaps in the existing literature. Eligibility Criteria: Eligible studies included children and adolescents aged 0–18 years diagnosed with ischemic stroke and receiving rehabilitation or therapeutic intervention. Studies addressing the timing, intensity, and effects of physiotherapy, occupational therapy, speech and language therapy, neuropsychological intervention, neuromodulation, or multidisciplinary rehabilitation were considered for inclusion. Sources of Evidence: A structured literature search was conducted in PubMed/MEDLINE, Scopus, Web of Science, the Cochrane Library, and Google Scholar for studies published between 2000 and January 2025. Charting Methods: Data were extracted using a standardized charting form and synthesized narratively because of substantial heterogeneity in study design, populations, interventions, and outcome measures. Results: Twenty-one sources met the inclusion criteria. Direct evidence specifically addressing early rehabilitation after pediatric ischemic stroke was limited and consisted primarily of observational studies. A substantial proportion of the available evidence was indirect, originating from studies of perinatal stroke, unilateral brain injury, cerebral palsy, and related pediatric neurorehabilitation populations, as well as clinical guidelines and expert consensus documents. The available evidence suggests potential benefits across motor, cognitive, communication, and functional domains, although the strength and directness of evidence varied substantially. Several studies identified the early post-stroke period as a potentially important window for neuroplasticity, while family involvement, individualized treatment planning, and interdisciplinary care were consistently highlighted as important components of rehabilitation. Evidence supporting neuromodulation techniques remained preliminary and was largely limited to selected pediatric populations. Conclusions: The available evidence, although heterogeneous and largely indirect, suggests that early coordinated and multidisciplinary rehabilitation may be beneficial in pediatric ischemic stroke care. However, the current evidence base remains limited, and high-quality prospective studies are needed to establish standardized rehabilitation protocols and determine the optimal timing and intensity of therapeutic interventions. Full article

Advanced cyber threats increasingly exploit volatile memory to execute malicious payloads without touching persistent storage, rendering traditional disk-centric forensic tools insufficient for comprehensive digital investigations. This paper presents Forensic-BERT, an AI-driven forensic framework that automatically extracts and classifies potentially relevant artifacts from unstructured memory dumps across heterogeneous operating environments. The framework combines byte-boundary-preserving Hex-to-ASCII conversion, sliding-window Shannon entropy filtering ($H>7.2$ bits per byte, 256-byte windows) to isolate high-probability artifact regions, and a binary-aware WordPiece tokenizer extended with 2048 domain-specific tokens covering hexadecimal byte patterns, Windows API names, and Linux system-call sequences. These components feed a transformer-based classifier fine-tuned from bert-base-uncased (110 M parameters) on memory-derived text, with sliding-window inference and majority-vote aggregation for large images. A SHAP DeepExplainer module and averaged 12-head attention heatmaps provide transparent, analyst-accessible explanations for classification decisions. We evaluate the framework on a multi-source corpus of 735 labeled memory segments drawn from 197 distinct images across four independent collections, MemLabs, the DARPA Transparent Computing program, Digital Corpora, and live sandbox execution traces from Any.run and Joe Sandbox, spanning Windows XP through Windows 11, Ubuntu Linux 16.04/18.04, and FreeBSD. Source-stratified five-fold cross-validation yields an overall F1-score of $0.92\pm 0.02$ and AUC-ROC of $0.95\pm 0.01$ (95% CI). Forensic-BERT outperforms all six baselines, Volatility with YARA rules (F1 $=0.71$), Random Forest (F1 $=0.82$), BiLSTM with GloVe embeddings (F1 $=0.85$), MRm-DLDet (F1 $=0.87$), SPECTRE (F1 $=0.89$), and SecBERT (F1 $=0.90$), with every pairwise difference statistically significant under the McNemar test with Bonferroni correction. Explainability quality is independently confirmed by a Spearman rank correlation of $\rho =0.81$ between model SHAP token rankings and expert forensic-indicator rankings and by a System Usability Scale score of 73.2 among certified examiners. The complete pipeline processes 512 MB memory images in 7.5–10.2 s (GPU) or 38–52 s (CPU-only), scaling to 4 GB images with near-linear throughput. These results indicate that, on the corpus evaluated here, combining domain-adapted NLP preprocessing, transformer-based sequence modeling, and quantified explainability can improve the effectiveness and usability of analyst decision support and investigative triage for volatile memory analysis. Full article

Capacitated vehicle routing with time windows (CVRPTW) is a natural target for coherent Ising machines (CIMs), but a direct multi-vehicle arc encoding scales as $O\left(m{N}^2\right)$ and exceeds the variable budget of current CIM-compatible systems. We argue the bottleneck is encoding density, not expressiveness, and present LSQ, a hardware-aware sparse Quadratic Unconstrained Binary Optimization (QUBO) framework that decouples CVRPTW into a compact customer-to-route assignment QUBO and a classical intra-route ordering step under a soft no-wait service convention. LKH candidate edges compress the per-route edge space from $O\left(N^2\right)$ to $O\left(KN\right)$, and a per-route dynamic-penalty subroutine encodes time-window sensitivities as binary variables in a round-wise outer loop. On a six-vehicle, 51-node reference instance curated from long-term operational data, LSQ shrinks the maximum single-submission QUBO from 15,300 arc variables to 342 logicalQUBOvariables (∼45× compression), cuts travel time by 22.9% (74 vs. 96), and cuts route duration by 11.2% (174 vs. 196) against an OR-Tools soft-window baseline at the same fleet size. OR-Tools retains an advantage on raw time-window penalty (1600 vs. 3540) and runtime; under the scalar operational cost ${\sum }_{k}T\left({\pi }_k\right)+{\sum }_i{\ell }_i\left({\tau }_i\right)$, OR-Tools is therefore the better single-objective solver, and the comparison is a multi-objective trade-off rather than a scalar dominance claim. Ablations confirm that the LKH prior recovers Held–Karp on a 15-customer TSP at 53 vs. 120 variables and that the dynamic-penalty encoding reduces compressed time-window loss by 15.25% at constant travel. All hardware claims refer to QUBO sizing on a Kaiwu/CIM-compatible backend, not physical CIM execution. Full article

Predicting near-future monetization in virtual livestreaming remains methodologically challenging because paid-support events are sparse, temporally dependent, and vulnerable to leakage under inappropriate evaluation designs. This study develops a leakage-guarded, window-based machine-learning framework for predicting next-window Superchat occurrence from VTuber live-chat dynamics. Public VTuber live-chat and Superchat logs were reconstructed into non-overlapping five-minute windows, and features were organized into audience activity, member composition, message intensity, donation-state information, and short-horizon dynamic groups. To reduce optimistic bias, the primary evaluation used video-level grouped splitting and compared a strict setting that excluded direct current-window donation-state variables with an extended donation-state-aware setting. HistGradientBoosting achieved the strongest performance. In the strict setting, it reached PR-AUC = 0.899, ROC-AUC = 0.920, F1 = 0.822, and Brier score = 0.171, while the extended setting produced only modest additional gains. Additional zero-chat sensitivity, repeated grouped split, channel-level robustness, graph-proxy baseline, feature-ablation, and calibration analyses supported the stability and interpretability of the framework. The results suggest that next-window Superchat occurrence can be predicted from participation breadth, chat activity, message intensity, and temporally shifted behavioral dynamics under leakage-aware evaluation. Full article