Search Results (295)

The effective allocation of Intensive Care Unit (ICU) beds is critical for balancing timely access across patient groups of differing clinical urgency, particularly during demand surges. This paper extends the threshold-based loss-queuing framework to three patient classes, emergency, semi-urgent, and elective, governed by a two-tier coordinate reservation policy with thresholds $(k_1,k_2)$. Under Poisson arrivals, we derive an exact product form steady-state distribution and closed-form blocking probabilities for all three classes; under Interrupted Poisson Process (IPP) arrivals, we construct a block-tridiagonal Markov chain over the full $(i,j,l,\xi )$ state space and obtain dimensionally consistent blocking formulas via a lifted rate matrix. Fitted to the publicly available MIMIC-IV Medical ICU dataset, the IPP captures bursty emergency arrival patterns with mean and variance deviations below $0.1\%$. A comprehensive parameter sensitivity sweep over traffic utilization $\rho \in [0.40,0.95]$ and burstiness index $z\in [1.0,5.0]$ identifies three distinct operating regimes and yields policy recommendation charts for direct clinical use. An adaptive optimization framework selects $(k_1^{*},k_2^{*})$ to minimize a weighted blocking loss subject to fairness constraints, achieving a Jain fairness index above $0.999$ throughout the high-load region. Analytical predictions are validated against discrete-event simulation with a $100\%$ pass rate at a ±1.5 percentage-point criterion. We further demonstrate that the blocking probabilities for all three classes are insensitive to the LoS distribution beyond its mean across six service-time distributions spanning coefficients of variation from $0.45$ to $2.24$, with a $100\%$ pass rate across all 72 (class, distribution, setting) combinations, broadening the model’s applicability to diverse real-world scenarios. The findings provide actionable guidance for ICU managers in determining fair and efficient three-tier bed reservation thresholds. Full article

High-speed train aerodynamics have mainly been improved by passive design methods, such as streamlined noses, local fairings, and surface smoothing. These methods have achieved clear benefits, but several important aerodynamic problems remain difficult to solve by geometry optimization alone. Open-air drag is still affected by tail flow separation, base-pressure recovery, and disturbances around bogies and the underbody; crosswind safety is influenced by unsteady leeward-side separation and wake asymmetry; slipstream behavior depends on wake vortices, boundary-layer development, and complex near-ground underbody flow; and tunnel-related pressure transients arise from compression-wave generation, propagation, and reflection. These coupled effects mean that one fixed train shape cannot perform optimally in all operating conditions. For this reason, this review proposes that active flow control (AFC) should not be regarded only as a drag-reduction or stability-improvement technique for high-speed trains. Instead, it should be understood as a mission-adaptive aerodynamic control framework, in which different control actions are used for different operating scenarios. This paper first clarifies that passive optimization is increasingly subject to diminishing returns under multi-objective and engineering constraints. It then reviews AFC studies on drag reduction, base-pressure recovery, wake and slipstream control, underbody flow conditioning, crosswind mitigation, and tunnel pressure-wave suppression. Related AFC studies on bluff bodies, road vehicles, and other separated flows are included only when their physical relevance to trains is clear. The review further distinguishes gross aerodynamic improvement from net energy gain and identifies actuator power, durability, maintainability, acoustic impact, validation level, and full-scale transferability as decisive feasibility factors. Current research is still dominated by open-loop numerical studies with simplified actuation. Future work should therefore move toward multi-objective, closed-loop, energy-aware, sensor–actuator-integrated, and explainable machine-learning-assisted AFC. The main message is that the next step in train aerodynamics is not simply a better fixed shape, but a control-enabled train that can selectively redistribute aerodynamic authority across its mission profile. Full article

Prosumer communities, aggregations of residential and commercial entities equipped with distributed energy resources (DER), including photovoltaic systems, battery storage, and flexible loads, are emerging as critical organizational units in decarbonising smart grid architectures. Managing these communities effectively requires balancing economic efficiency with equity, autonomy, and environmental sustainability, objectives that conventional centralized control methods and existing multi-agent reinforcement learning (MARL) implementations fail to address simultaneously. This article proposes a value-aligned hierarchical multi-agent reinforcement learning (VA-HMARL) framework as a formally unified architecture that embeds equity (Jain’s Fairness Index J ≥ 0.90), individual autonomy, and carbon sustainability as hard constraints within the MARL reward structure. The framework integrates: a multi-objective Value Alignment Module (VAM) combining economic, fairness, sustainability, and comfort objectives; attention-based implicit coordination for scalable agent interaction; and differentially private federated policy aggregation (ε = 1.0, δ = 10⁻⁵) for GDPR-compliant collaborative learning. Simulation on a 20-prosumer community modelled on the IEEE 33-bus feeder over 10 Monte Carlo runs (300 episodes each) demonstrates: a 6.2% energy cost reduction versus the Rule-Based baseline (p = 0.0004); a Jain’s Fairness Index of 0.912 ± 0.031 at policy convergence (final 50 episodes), satisfying the J ≥ 0.90 community equity floor; and an 18.0% reduction in CO₂ emissions. The economic efficiency trade-off relative to performance-optimized MARL baselines is limited to 2.4%, within the 5% design target. These results establish VA-HMARL as a technically feasible and ethically grounded paradigm for autonomous decentralized energy governance. Full article

The electrification of field service fleets introduces complex constraints: shift limits, overtime fairness, and battery–range feasibility. This paper proposes the Multi-Shift Single Electric Vehicle Routing Problem under Possibilistic Uncertainty (MS-SEVRP-PU), a formulation focused on a single-vehicle multi-shift planning unit and capturing imprecise travel/service times and state-of-charge dynamics. Travel durations and energy consumption are modelled as triangular fuzzy numbers to reflect expert knowledge when probabilistic data is limited. A closed-form credibility function evaluates overtime risk, while an Ordered Weighted Averaging (OWA) aggregation of per-shift risks ensures fairness by discouraging systematic overload on specific shifts. To solve this multi-objective problem, we develop a Pareto-Conditioned Transformer with risk-aware and battery-conscious large neighbourhood search (PCT-RABLNS), combining a preference-conditioned attention policy with targeted local search. Computational experiments on calibrated municipal maintenance case studies indicate that PCT-RABLNS improves hypervolume by 2–5% over strong baselines and reduces maximum shift overtime risk by 15–25%, with a marginal makespan overhead of only 1–3%. The results demonstrate that the proposed framework is a promising decision-support approach for energy-aware, risk-fair, and operationally compliant planning of single-vehicle, multi-shift electric service operations, jointly integrating multi-shift routing, fuzzy uncertainty, and preference-conditioned reinforcement learning. The paper also discusses how the framework can be extended to multi-vehicle settings. Full article

Artificial Intelligence (AI) and Digital Twins (DTs) can support the digital and energy transition in the construction sector; however, their application to the built environment presents both opportunities and limitations. This study aims to give a critical perspective on the topic analyzing the related key challenges, including error assessment, model interpretability, data availability, cybersecurity risks, organizational constraints, and lifecycle costs. Where AI is nowadays developed as a context-dependent tool set, it is most effective when embedded within integrated socio-technical systems rather than adopted as a universal solution. Instead, DTs can be intended as an enabling framework, integrating AI, Internet of Things (IoT), Big Data, and Building Management Systems (BMS) to enhance energy performance, indoor environmental quality, safety, maintenance, and decision-making at both building and urban scales. The direction international research on these topics is facing is clear as evidenced by the wide number of research papers published. The future of these technologies moves towards a simulative approach oriented towards the sustainable and fair development goals and will bring a broad transformation of the building environment where they are ever more integrated into each social and technical aspect. The work is supported by a case study developed at Sapienza University of Rome founded by the Italian National Recovery and Resilience Plan within Flagship Project 2 (FP2), “Energy Transition and Digital Transition in Urban Regeneration and Construction,” of the Rome Technopole ecosystem. Full article

In next-generation wireless communication systems, spectrum efficiency can be realized through the integration of hybrid beamforming (HBF) and non-orthogonal multiple access (NOMA). To maximize the synergy between these two technologies, it is essential to accurately cluster users within beams. Most existing studies on clustering overlook practical constraints and assume perfect channel state information (CSI). However, obtaining full CSI is impractical in realistic environments due to high feedback overhead and potential CSI errors. To address these challenges, this paper adopts an opportunistic beamforming (OBF) framework based on a partial CSI environment. The OBF facilitates channel estimation and HBF precoder design using only signal-to-interference-plus-noise ratio (SINR) feedback. Subsequently, clustering and power allocation (PA) are performed utilizing the feedback SINR from OBF without requiring additional feedback information. While conventional NOMA focuses on maximizing either throughput or fairness, this paper proposes a scheme that selects users with high SINR to maximize system throughput while minimizing the throughput disparity among users to enhance fairness. Furthermore, a power allocation method that satisfies the minimum successive interference cancellation (SIC) power requirement is employed to ensure stable decoding. Simulation results demonstrate that the proposed clustering scheme enhances the sum-rate compared to conventional SINR-based clustering methods while maintaining fairness. Consequently, this study suggests a promising approach to improving NOMA performance in practical partial CSI environments. Full article

Ensuring fairness in machine learning while maintaining predictive performance remains a fundamental challenge in data science. Most fairness-aware learning approaches rely on fixed penalty scalarization or static multi-objective formulations, which often lead to unstable trade-offs and sensitivity to manually tuned hyperparameters. In this paper, we propose SAFEA (Self-Adaptive Fairness Entropy Algorithm), a novel evolutionary optimization framework that dynamically regulates the fairness–accuracy trade-off using inequality-aware feedback mechanisms. SAFEA introduces two complementary measures: the Fairness Entropy Index (FEI), which captures the dispersion of group-level fairness violations, and the Gini Fairness Index, which quantifies disparity in prediction errors across protected groups. These measures guide an adaptive penalty update rule that autonomously adjusts the fairness coefficient during the evolutionary search process, eliminating the need for manual tuning. Theoretical analysis establishes boundedness and stability of the adaptive penalty under mild assumptions and discusses convergence properties under Lipschitz-continuous objectives. Experimental evaluation on benchmark datasets (Adult Income, COMPAS, and German Credit) demonstrates that SAFEA improves hypervolume by up to 12.4% compared to NSGA-II fairness formulations, reduces demographic parity difference by 18–25% relative to static penalty evolutionary methods, and achieves up to 3.1% higher F1-score than adversarial debiasing approaches while maintaining competitive accuracy. These results indicate that entropy-guided adaptive regulation leads to smoother fairness convergence and better Pareto front coverage. The proposed framework bridges inequality theory and evolutionary multi-objective optimization, providing a scalable and effective solution for fairness-aware learning in high-stakes applications. Full article

Background: An accurate D’Amico risk stratification is mandatory for prostate cancer (PCa) management. The purpose of this proof-of-concept study was to establish a methodological framework for integrating validated clinical nomograms with strict data-quality governance in order to generate reliable artificial neural networks (ANNs), even when the sample is small. Methods: We performed a retrospective analysis of a curated cohort of 49 patients from one centre. A multilayer perceptron (MLP) was trained using 11 variables, including the ISUP biopsy grade and Briganti nomogram. Model development was guided by a proactive data-quality protocol based on FAIR principles—the DQG-AI framework (data quality governance for AI-readiness, developed at Clínica Universidad de Navarra)—with stringent checks for accuracy, consistency and validity to ensure data were “AI-ready”. A sensitivity analysis was conducted on three data partitioning scenarios (20/80, 34/66 and 39/61). Results: From a starting pool of 76 patients, the DQG-AI framework was applied to create a highly selected cohort of 49 patients. A multilayer perceptron (MLP) trained on this “AI-ready” dataset achieved, on the 20/80 configuration, mathematically perfect discrimination (AUC 1.000; 100% accuracy) for High vs. Intermediate risk groups on a very small refined internal test set (N = 9), a figure we interpret as a methodological artefact of the curated dataset and validation constraints rather than as an indicator of true model performance. This complete accuracy is not, however, presented as evidence of generalizable clinical utility: it is a best-case figure obtained on a single, very small test subset (N = 9) after necessary validation-related exclusions, and the wide confidence interval (66.4–100%), together with the software-driven removal of test cases carrying factor levels absent from the training set (detailed in the Methods section), explicitly preclude any inference about real-world performance. Accordingly, the deliverable of this proof-of-concept study is the DQG-AI framework itself, not the model’s reported accuracy. Conclusions: The main contribution of this proof-of-concept study is the effective illustration of the DQG-AI framework as a strict, repeatable approach for producing “AI-ready” urological datasets. Although the MLP demonstrated a robust internal signal for risk discrimination, its flawless accuracy is an ideal, non-generalizable situation. The most important deliverable that needs external validation is the DQG-AI framework, not the model’s performance metrics. A pre-specified three-phase multi-institutional validation roadmap (single-centre cohort expansion → within-system between-site validation → Spanish multi-centre external validation), with a minimum target of ~220 evaluable patients derived from a 10-events-per-predictor floor, is provided to operationalise this external validation. Full article

The rapid growth of the Internet of Things (IoT) has introduced significant cybersecurity challenges due to the heterogeneity, scale, and limited protection capability of connected devices. Although machine learning has been widely adopted for IoT intrusion detection, many existing studies still rely primarily on closed-world evaluation settings, unequal baseline comparison budgets, fixed decision thresholds, and limited integration of explainability into model assessment. To address these issues, this paper proposes an explainable XGBoost-based framework for IoT attack detection with unseen attack family evaluation using the large-scale CICIoT2023 dataset. In the proposed framework, IoT traffic is formulated as a binary classification task that distinguishes benign from malicious flows. The study integrates two complementary evaluation protocols: (1) closed-world stratified 10-fold cross-validation for in-distribution performance assessment and (2) unseen attack family evaluation, in which one malicious family is excluded from training and used only for testing under a zero-day-like but single-dataset condition. A fair-budget experimental design is adopted to compare seven representative models under the same training budget, including default XGBoost, optimized XGBoost, Random Forest, LightGBM, CatBoost, Logistic Regression, and a simple multilayer perceptron. To improve reproducibility and operational validity, the revised framework further reports the sampling strategy, split-overlap audit, XGBoost hyperparameter search protocol, repeated unseen-family evaluation, validation-based threshold calibration under fixed-FAR constraints, cost-sensitive threshold analysis, and XGBoost-native SHapley Additive exPlanations (SHAP) compatible feature contribution analysis. The closed-world results show that tree-based ensemble methods clearly outperform the linear and shallow neural baselines. Random Forest achieves the highest closed-world macro-F1 of 0.9713, followed by LightGBM with 0.9602 and optimized XGBoost with 0.9566. In the fair-budget unseen-family setting under the default threshold, Random Forest again obtains the highest mean macro-F1 of 0.8433 and the lowest false negative rate (FNR) of 0.0712, but it also produces a substantially higher false alarm rate (FAR = 0.0536). By contrast, optimized XGBoost provides a lower-FAR default operating point, achieving a mean macro-F1 of 0.8194, Matthews correlation coefficient (MCC) of 0.7067, FAR of 0.0086, and FNR of 0.2996. Repeated unseen-family experiments over five random seeds confirm the same trade-off: Random Forest provides stronger recall-oriented detection, whereas optimized XGBoost provides a lower-FAR default operating point. After validation-based threshold calibration at an approximate FAR target of 0.01, Random Forest achieves the strongest calibrated recall-oriented performance, with macro-F1 of 0.8754, MCC of 0.7757, FNR of 0.2000, and attack recall of 0.8000. Optimized XGBoost remains competitive at the same FAR target, with macro-F1 of 0.8323, MCC of 0.7193, FNR of 0.2760, and attack recall of 0.7240. The explainability analysis indicates that the optimized XGBoost detector relies mainly on TCP control-flag, temporal, and packet-statistical features, with rst_count, IAT, urg_count, Tot size, Number, Header_Length, and Magnitude among the most influential variables. Local contribution tables for representative true-positive, false-positive, false-negative, and true-negative cases further improve the readability of the explanation results and confirm that native pred_contribs reconstructs the model margin with negligible numerical error. Overall, the results show that the most appropriate model depends on the deployment objective: Random Forest is preferable when minimizing missed attacks under a calibrated FAR constraint is prioritized, whereas optimized XGBoost remains a strong primary model for an explainable low-FAR XGBoost-based framework that emphasizes scalability, operational conservativeness, and native contribution-based interpretation. Full article

Power electronic systems are often engineered through a sequential–iterative workflow in which hardware parameters are initially sized from steady-state, ripple, thermal, and electromagnetic-compatibility constraints, and controllers are subsequently tuned to satisfy dynamic and closed-loop performance requirements. While converters are inherently designed for closed-loop operation, increasing power density, uncertainty, and distributed interaction make the underlying design process resemble a strategic interplay among multiple decision-makers, including hardware designers, control algorithms, loads, disturbances, and manufacturing constraints. This paper develops a unifying game-theoretic perspective on the optimal design and control of power electronic systems. Classical concepts—such as robust control, worst-case design, droop-based load sharing, and tolerance allocation—are reinterpreted as equilibrium solutions of zero-sum, Stackelberg, non-cooperative, or cooperative games. Beyond a conceptual taxonomy, two illustrative simulation case studies are provided: (i) a Stackelberg hardware–controller co-design of a buck converter, demonstrating simultaneous passive-component reduction and improved transient performance relative to a conservative sequential design; and (ii) a droop-controlled parallel-converter example contrasting Nash and cooperative equilibria, explicitly quantifying trade-offs between bus-voltage regulation, current-sharing fairness, and conduction losses. By framing power electronic design and control as interacting strategic processes rather than isolated optimization stages, the paper aims to show that game theory can serve as a structured and practically interpretable framework for distributed and uncertainty-aware power electronic systems. Full article

The public deployment of electric vehicle charging stations must simultaneously balance construction economics, user accessibility, queueing pressure, feeder security, tail risk under demand uncertainty, and spatial fairness. These criteria are strongly coupled, yet most existing studies either rely on static optimization with limited behavioral realism or use multi-agent reinforcement learning for short-term charging operation rather than for long-term siting. This paper proposes a preference-conditioned multi-agent deep deterministic policy gradient (PC-MADDPG) framework for the urban charging station siting problem in a coupled traffic–distribution environment. Candidate charging sites are modeled as cooperative agents under centralized training and decentralized execution. Each agent outputs a continuous pile-allocation action, which is repaired into an integer expansion plan under a budget constraint. The environment evaluates each plan through attraction-based demand assignment, queue approximation, LinDistFlow-style feeder analysis, and a six-objective performance vector, including annual net cost, travel burden, service inconvenience, grid penalty, CVaR of unmet charging demand, and equity loss. On a reproducible benchmark with 12 demand zones, 10 candidate sites, an 11-bus radial feeder, and 16 stochastic daily scenarios, the proposed framework generates a non-dominated archive with 42 unique feasible plans. A representative PC-MADDPG solution opens 5 of 10 candidate sites and installs 20 fast-charging piles, achieving 99.88% mean demand coverage with an annual profit of 2.083 M$ and a maximum line utilization of 0.999. Relative to the NoGrid ablation, the selected full model reduces grid penalty by 23.87% and equity Gini by 51.08%, with only a 0.35% profit concession. Relative to the NoRisk ablation, the CVaR of unmet demand is lowered by 69.70%. Compared with a demand-greedy baseline, the proposed method reduces grid penalty by 11.72% and equity Gini by 25.19% while preserving similar demand coverage. These results provide proof-of-concept evidence, on a reproducible coupled benchmark, that preference-conditioned multi-agent learning can serve as a practical many-objective siting engine for charging-infrastructure planning when coupled traffic and feeder constraints are explicitly modeled. Full article

We present an updated overview of the symmetry-preserving contact interaction model in hadronic physics, which was developed a little over a decade ago to describe the mass spectrum and internal structure of mesons and diquarks composed of light and heavy quarks. Over the years, the contact interaction model has evolved into a framework capable of treating both ground and excited states, providing a simple yet consistent approach to nonperturbative QCD. In this review, we examine the mass spectrum and elastic form factors of forty mesons with different spins and parities, together with their corresponding diquark partners. Importantly, we update the comparison of contact interaction predictions using recent results from the literature, offering a fresh perspective on the model’s performance, strengths, and limitations. The analysis presented here refines previous conclusions and supports the contact interaction model as a practical tool for hadron structure studies, with potential applications to baryons and multiquark states. We also present comparisons with other theoretical models and approaches, including lattice quantum chromodynamics, and comment on future prospects in view of ongoing and planned experimental programs regarding hadron structure. In particular, forthcoming measurements at FAIR together with future studies at Jefferson Lab and the Electron Ion Collider are expected to provide key insights into hadron structure, with FAIR offering indirect constraints via hadron spectroscopy, hadronic interactions, and in-medium properties; high-precision data on meson structure and form factors from Jefferson Lab and the Electron Ion Collider will provide valuable benchmarks with which to confront predictions based on the contact interaction model. Full article

The inherent redundancy in vehicle sensor data, coupled with constrained onboard resources and stringent latency requirements, renders traditional bit-oriented transmission paradigms inefficient for autonomous-driving perception tasks. Semantic communication offers a promising direction by shifting the focus from bit-level fidelity to task-level information delivery. In this paper, we propose a unified framework that integrates hierarchical transmission and online scheduling for Internet of Vehicles (IoV)-oriented collaborative perception. The proposed hierarchy separates information into two complementary layers: a coarse metadata layer (object bounding boxes) for latency-critical awareness, and fine-grained visual semantics (multi-scale region-of-interest (ROI) patches) for perception-intensive tasks. We formulate an online scheduling problem that jointly exploits Age of Information (AoI) and Channel State Information (CSI) to dynamically decide what to transmit and at what fidelity under per-frame budget constraints. To address cross-scheme fairness, we report resource utilization under a fixed kbps/fps physical budget and evaluate robustness using a combination of a lightweight task-proxy metric and COCO-style Average Recall (AR100) under ROI-only evaluation. The hierarchical transmission architecture, combined with AoI awareness, reduces global semantic staleness by approximately 78%. The Lyapunov-based online scheduler enables intelligent, signal-to-noise ratio (SNR)-adaptive switching between coarse and fine semantic levels, ensuring robust perception under varying channel quality. Under strict physical-budget constraints and unreliable channel conditions, joint source-channel coding (JSCC) exhibits significantly stronger task robustness than conventional schemes: at 0 dB SNR, the task-proxy detection rate improves by nearly 47 percentage points over the uncoded baseline. Full article

Recurring socio-temporal events, such as Ramadan in Middle Eastern cities, introduce pronounced non-stationarity in urban traffic demand. During these periods, daytime traffic volumes typically decline, while congestion becomes more severe in the evening around the Iftar (fast-breaking) period and persists into late-night hours, making conventional fixed-time signal plans less effective. An additional challenge is that demand is not only time-varying, but also unevenly distributed across competing movements: attempts to prioritize high-volume phases can inadvertently cause excessive delays—or even starvation—on lower-demand approaches. To address these issues, this study presents an adaptive, regime-aware traffic signal control framework that combines predictive modeling with constrained optimization. Short-term phase-level delays are forecast using Long Short-Term Memory (LSTM) models, and a Model Predictive Control (MPC) scheme then determines the green time allocation at each control cycle through a receding-horizon strategy. The optimization explicitly represents phase interactions by including constraints that prevent excessive delay in competing movements, thereby yielding a balanced and operationally realistic control policy. The approach is validated with one-minute-resolution TomTom delay data from a signalized intersection in Jeddah, Saudi Arabia, covering both Normal and Ramadan conditions. The LSTM models show stable predictive performance, achieving root mean square errors (RMSEs) of 19.8 s under Normal conditions and 17.1 s during Ramadan. In general, the results show that the proposed framework cuts total intersection delay by about 0.3% to 2.8% compared to standard control strategies. Even though these total-delay improvements are small, they come with big drops in delay for lower-demand phases (about 12–20%) and keep the delay increases for higher-demand phases under control. This shows that the method makes the whole process more efficient by fairly spreading out the delay instead of just making one phase better on its own. The results show that combining forecasting with constrained optimization is a strong and useful way to handle changing traffic demand. This is especially true during times of high demand when flexibility, stability, and fairness across movements are all important. Full article

Managing blood glucose in type 1 diabetes (T1D) remains a daily clinical challenge, and accurate short-term prediction of glucose levels can meaningfully improve insulin dosing decisions while reducing the risk of dangerous hypoglycaemic episodes. Although numerous machine learning approaches have been proposed for this task, comparing their relative merits is difficult because published studies differ widely in datasets, preprocessing choices, and evaluation criteria. In this work, we address this research gap by benchmarking ten machine learning methods—from a naïve persistence baseline through classical linear regressors, gradient-boosted ensembles, and recurrent neural networks to a novel hybrid that couples LightGBM with stochastic differential equation (SDE)-based glucose–insulin simulation—on two multi-patient datasets comprising 34 T1D subjects, across prediction horizons of 15, 30, 60, and 120 min. Every method is trained and tested under identical preprocessing and temporal splitting conditions to ensure a fair comparison. The proposed Hybrid LightGBM-SDE model consistently outperforms all alternatives, recording RMSE values of 22.42 mg/dL at 15 min, 28.74 mg/dL at 30 min, 33.89 mg/dL at 60 min, and 37.22 mg/dL at 120 min—an improvement of between 13.6% and 27.0% relative to standalone LightGBM. At the clinically important 30 min horizon, 99.7% of predictions lie within the acceptable A and B zones of the Clarke Error Grid. Wilcoxon signed-rank tests confirm that performance differences are statistically significant (p < 10⁻¹⁰), and SHAP-based analysis shows that the SDE-derived simulation features are among the most influential predictors, especially at longer horizons. All source code and evaluation scripts are publicly released to support reproducibility. Due to temporary data access constraints, all experiments reported here use physics-based synthetic datasets generated from the Bergman minimal model, replicating the structural properties of the D1NAMO and HUPA-UCM collections; validation on the original clinical recordings is planned. Among the two synthetic datasets, the D1NAMO-equivalent cohort (nine patients) proves more challenging, with systematically higher per-patient RMSE variance. The clinically acceptable prediction accuracy at the 30 min horizon (99.7% in Clarke zones A + B) suggests potential for integration into insulin dosing decision-support systems. Full article