Search Results (51)

Field-based phenotyping of water-related traits for precision irrigation in tropical agroecosystems poses a persistent methodological challenge, driven by high climatic variability and the complex water-use physiology of Crassulacean Acid Metabolism (CAM) crops such as pineapple (Ananas comosus var. MD2). We developed and validated a Physics-Informed Machine Learning (PIML) framework that integrates high-resolution UAV multispectral imagery, IoT-based microclimatic records, and a mechanistic soil water balance based on the FAO-56 Penman–Monteith standard to predict plot-scale soil moisture depletion as a proxy of plant water status. A six-month field campaign (March–August 2022) across 25 georeferenced commercial pineapple plots in the Colombian Orinoquia piedmont yielded a spatiotemporally balanced dataset of $N=150$ observations. Soil-adjusted vegetation indices (OSAVI, MSAVI) outperformed standard NDVI for capturing water-related canopy traits, effectively decoupling spectral responses from substrate noise. A Gradient Boosting regressor achieved $R^2=0.842$ and $RMSE=0.0705$ on a normalized target scale, corresponding to a $7.05\%$ error over the prediction range, while the traffic-light Decision Support System (DSS) for irrigation scheduling reached $91.1\%$ accuracy (Cohen’s Kappa $=0.91$). Incorporating daily soil moisture depletion as a mechanistic feature improved predictive accuracy over a spectral-only baseline ($\mathrm{\Delta }{R}^2=+0.052$) and anchored predictions within a physically consistent framework based on the FAO-56 water balance, with no false negatives observed for water deficit detection in the hold-out validation set. This framework advances high-throughput, population-scale phenotyping of water-related traits in open-canopy CAM crops, establishing a transferable methodology for operational precision irrigation under tropical savanna conditions. 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

Food delivery logistics sustainability includes environmental impact, economic efficiency, and service quality. Traditional logistics models mainly rely on fixed “pickup buffer” policies (such as a set 10 min wait). These systems do not account for the changing nature of restaurant operations and delivery conditions, leading to higher operating costs, driver idle time, and poorer food quality. To move delivery systems from reactive decision-making to proactive, dynamically forecasted operations, an adaptive control mechanism is needed. In on-demand food delivery, this offers a clear path to sustainability through better dispatch accuracy, order prep, and pickup coordination. To resolve these bottlenecks, this study examines how a smart logistics framework based on a dynamic Gradient Boosting Regressor (GBR) and policy-sensitive GBR can provide more accurate estimates of drivers’ waiting times in light of contextual factors such as rush hour, time of day, and operational constraints. In last-mile food delivery, the proposed method aims to reduce operational costs, improve scheduling effectiveness, and maximize resource utilization by moving beyond static, predefined waiting periods to adaptive, context-aware decisions. The developed framework analyzes a proprietary dataset of 368,250 instant orders from a major Saudi Arabian logistics provider to evaluate the efficacy of static thresholds versus a proposed predictive, dynamic machine-learning-based approach. After rigorous data cleaning and temporal-logic adjustments, a “High-Fidelity Ground-Truth” subset of 1842 verified orders is used to simulate policy performance. This 99.5% reduction is necessitated by the widespread absence of the “Order Ready” timestamp in operational logs, which is the critical target variable for supervised learning; comparative analysis confirms the subset remains representative of the parent population’s spatiotemporal dynamics. The baseline analysis reveals severe inefficiencies in the static model, with a 61.67% violation rate for driver wait times, particularly in Riyadh ($p<0.001$) and during late-night operations. The simulation results demonstrate that the dynamic policy reduces the “Buffer Miss Rate” (premature driver arrivals) from 59.08% to 7.32%, resulting in a 68.5% reduction in total operational waste costs. Full article

Bike sharing is an established component of urban mobility infrastructure, offering a low-emission alternative to motorized transport for short trips in cities worldwide. Accurate demand forecasting is essential for efficient system operation: it enables better bike redistribution, reduces user wait times, and lowers the operational costs associated with rebalancing. This study evaluated multiple ensemble strategies for hourly bike-sharing demand prediction, comparing bagging methods (Random Forest, Extra Trees), boosting methods (AdaBoost, Gradient Boosting Regressor, Histogram-based Gradient Boosting Regressor), and a Voting ensemble, while systematically investigating the impact of hyperparameter optimization. A repeated hold-out protocol was used, in which the dataset was randomly divided into 80% training and 20% test subsets across 10 random splits; 5-fold cross-validation was applied within each training fold exclusively for hyperparameter tuning, ensuring the test set remained unseen during model selection. Random Search and Bayesian Optimization were compared under identical budgets of 60 configurations per model. Results show that optimization substantially improves all models, with the most pronounced gains for AdaBoost (58% RMSE reduction) and Gradient Boosting Regressor (45% RMSE reduction). A Voting ensemble combining a Random Search-tuned Gradient Boosting Regressor and a Bayesian-optimized Histogram-based Gradient Boosting Regressor achieves the best overall performance (RMSE of 38.48, R² of 0.955) with the lowest variance among all repeated splits. Feature importance analysis confirms that hour of day and temperature are the dominant demand drivers, consistent with the operational patterns of urban bike-sharing systems. The performance difference between Random Search and Bayesian Optimization is negligible for most models, suggesting that well-designed search spaces allow simpler strategies to achieve competitive results. A controlled comparison conducted under identical experimental conditions shows that the Voting ensemble is statistically equivalent to XGBoost and nominally better than LightGBM, while CatBoost achieves a statistically significant advantage, highlighting it as a strong individual alternative. Full article

Daily equity return forecasting is challenging due to low signal-to-noise ratios, heavy-tailed innovations, and persistent distribution drift. We study one-step-ahead log-return prediction using daily market variables and return-based transformations. We propose a CNN–LightGBM hybrid that transfers a last-step CNN embedding to a gradient-boosted tree regressor through explicit embedding standardization, which stabilizes the representation interface for tree learning. To reduce train-to-evaluation mismatch under drift, we adopt split-wise, training-only standardization with a recency-aware fit-latest-W rule. Return-related predictors are anchored on a one-sided wavelet-denoised close series, while other market channels are retained in their original form to preserve episodic extremes. Experiments on NIFTY50 with walk-forward model selection show statistically reliable accuracy gains over Naive0 and competitive performance against representative deep sequence baselines, and the supplementary evaluations on HDFC and INDA provide additional out-of-sample evidence on these two assets under the same strictly chronological protocol. A long-or-cash decision rule based on the return forecasts yields positive risk-adjusted performance under realistic transaction-cost assumptions, supporting the practical relevance of the predictive signal. Full article

Motivation: Accurate spectral reconstruction of printed halftone images is essential for achieving high-fidelity color reproduction and robust color management across modern printing systems. However, traditional physics-based models, such as the Yule–Nielsen and Clapper–Yule formulations, rely on simplified empirical assumptions and often fail to capture the complex nonlinear and asymmetric interactions induced by multi-ink overlays and substrate light scattering. Meanwhile, existing data-driven approaches based on single learning models exhibit limited capability in modeling the complementary and symmetrical characteristics inherent in halftone structures, resulting in suboptimal prediction accuracy and generalization performance. Method: To address these limitations, we propose a Stacking Ensemble Spectral Prediction (SESP) framework. The proposed method adopts a two-layer stacking architecture that integrates heterogeneous base regressors, including Support Vector Regression (SVR), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost 3.0.3), with Ridge Regression employed as the meta-learner for optimal prediction aggregation. This ensemble design enables effective modeling of both halftone pattern symmetry and complex substrate scattering behavior. Results: Extensive experiments conducted on printed halftone image datasets demonstrate the superior performance of the proposed SESP framework. Compared with the best-performing reference method (PCA-IPSO-DNN), SESP achieves relative reductions in RMSE and CIEDE2000 of 12.8% and 6.8% under illuminant A, 9.5% and 6.9% under D50, and 12.2% and 7.2% under D65, respectively. In addition, SESP consistently outperforms traditional physics-based models, including Yule–Nielsen and Clapper–Yule, in terms of both spectral prediction accuracy and colorimetric fidelity. These results confirm the effectiveness of the proposed framework in modeling the intricate nonlinear and asymmetric relationships between CMYK halftone patterns and spectral reflectance. Full article

Accurate wheat yield prediction is essential for ensuring food security and sustainable resource management under the increasing challenges of climate change. This study investigates the integration of unmanned aerial vehicle (UAV)-based multispectral imaging and machine learning (ML) techniques to improve yield forecasting in European wheat cultivars. Field experiments were conducted on 400 sub-plots with varying NPK fertilization regimes and five wheat varieties, monitored across six phenological stages during the 2023 growing season in Vojvodina, Serbia. A DJI Phantom 4 Multispectral UAV collected high-resolution imagery, from which 65 vegetation indices were computed. Using PyCaret’s automated ML framework, 25 regression algorithms were evaluated for yield prediction. Ensemble models, particularly Random Forest, Extra Trees, Gradient Boosting, and LightGBM, consistently outperformed linear and kernel-based approaches. The highest prediction accuracy was achieved with the Random Forest Regressor during full flowering (BBCH 65–69), yielding an R² of 0.952 and an RMSE of 0.44 t/ha. Results highlight the temporal dynamics of model performance, with optimal predictions occurring during reproductive stages. The findings confirm that UAV-derived multispectral data, coupled with ensemble machine learning, provide a non-invasive, accurate, and computationally efficient method for yield forecasting. This framework has significant potential for supporting precision agriculture, enabling real-time decision-making, and enhancing the resilience of wheat production systems. Full article

There are almost no studies that investigate the flexural behavior of existing reinforced concrete (RC) beams with insufficient concrete strength using machine learning methods. This study investigates the flexural response of low-strength concrete (LSC) RC beams reinforced exclusively with steel rebars, focusing on the effectiveness of three different longitudinal reinforcement configurations. Nine beams, each measuring 150 × 200 × 1100 mm and cast with C10-grade low-strength concrete, were divided into three groups according to their reinforcement layout: Group 1 (L2L) with two Ø12 mm rebars, Group 2 (L3L) with three Ø12 mm rebars, and Group 3 (F10L3L) with three Ø10 mm rebars. All specimens were tested under three-point bending to evaluate their load–deflection characteristics and failure mechanisms. The experimental findings were compared with ML approaches. To enhance predictive understanding, several ML regression models were developed and trained using the experimental datasets. Among them, the Light Gradient Boosting, K Neighbors Regressor and Adaboost Regressor exhibited the best predictive performance, estimating beam deflections with R² values of 0.89, 0.90, 0.94, 0.74, 0.84, 0.64, 0.70, 0.82, and 0.72, respectively. The results highlight that the proposed ML models effectively capture the nonlinear flexural behavior of RC beams and that longitudinal reinforcement configuration plays a significant role in the flexural performance of low-strength concrete beams, providing valuable insights for both design and structural assessment. Full article

This study presents the first application of Machine Learning (ML) models to optimise Powder Bed Fusion using Laser Beam (PBF-LB) process parameters for H13 steel fabricated on a 350 °C preheated building platform. A total of 189 cylindrical specimens were produced for training and testing machine learning (ML) models using variable process parameters: laser power (250–350 W), scanning speed (1050–1300 mm/s), and hatch spacing (65–90 $\mathsf{\mu }$m). Eight ML models were investigated: 1. Support Vector Regression (SVR), 2. Kernel Ridge Regression (KRR), 3. Stochastic Gradient Descent Regressor, 4. Random Forest Regressor (RFR), 5. Extreme Gradient Boosting (XGBoost), 6. Extreme Gradient Boosting with limited depth (XGBoost LD), 7. Extra Trees Regressor (ETR) and 8. Light Gradient Boosting Machine (LightGBM). All models were trained using the Fast Library for Automated Machine Learning & Tuning (FLAML) framework to predict the relative density of the fabricated samples. Among these, the XGBoost model achieved the highest predictive accuracy, with a coefficient of determination $R^2=0.977$, mean absolute percentage error MAPE = 0.002, and mean absolute error MAE = 0.017. Experimental validation was conducted on 27 newly fabricated samples using ML predicted process parameters. Relative densities exceeding 99.6% of the theoretical value (7.76 g/cm³) for all models except XGBoost LD and KRR. The lowest MAE = 0.004 and the smallest difference between the ML-predicted and PBF-LB validated density were obtained for samples made with LightGBM-predicted parameters. Those samples exhibited a hardness of 604 ± 13 HV0.5, which increased to approximately 630 HV0.5 after tempering at 550 °C. The LightGBM optimised parameters were further applied to fabricate a part of a forging die incorporating internal through-cooling channels, demonstrating the efficacy of machine learning-guided optimisation in achieving dense, defect-free H13 components suitable for industrial applications. Full article

Accurately forecasting the operating temperature of lithium-ion batteries (LIBs) is essential for preventing thermal runaway, extending service life, and ensuring the safe operation of electric vehicles and stationary energy-storage systems. This work introduces a unified, physics-informed, and data-driven temperature-prediction framework that integrates mathematically governed preprocessing, electrothermal decomposition, and sequential deep learning architectures. The methodology systematically applies the governing relations to convert raw temperature measurements into trend, seasonal, and residual components, thereby isolating long-term thermal accumulation, reversible entropy-driven oscillations, and irreversible resistive heating. These physically interpretable signatures serve as structured inputs to machine learning and deep learning models trained on temporally segmented temperature sequences. Among all evaluated predictors, the Bidirectional Long Short-Term Memory (BiLSTM) network achieved the highest prediction fidelity, yielding an RMSE of 0.018 °C, a 35.7% improvement over the conventional Long Short-Term Memory (LSTM) (RMSE = 0.028 °C) due to its ability to simultaneously encode forward and backward temporal dependencies inherent in cyclic electrochemical operation. While CatBoost exhibited the strongest performance among classical regressors (RMSE = 0.022 °C), outperforming Random Forest, Gradient Boosting, Support Vector Regression, XGBoost, and LightGBM, it remained inferior to BiLSTM because it lacks the capacity to represent bidirectional electrothermal dynamics. This performance hierarchy confirms that LIB thermal evolution is not dictated solely by historical load sequences; it also depends on forthcoming cycling patterns and entropic interactions, which unidirectional and memoryless models cannot capture. The resulting hybrid physics-data-driven framework provides a reliable surrogate for real-time LIB thermal estimation and can be directly embedded within BMS to enable proactive intervention strategies such as predictive cooling activation, current derating, and early detection of hazardous thermal conditions. By coupling physics-based decomposition with deep sequential learning, this study establishes a validated foundation for next-generation LIB thermal-management platforms and identifies a clear trajectory for future work extending the methodology to module- and pack-level systems suitable for industrial deployment. Full article

Travel planning is a process that allows users to obtain maximum benefit from their time, cost and energy. When planning a route from one place to another, it is an important option to present alternative travel areas on the route. This study proposes a travel route planning (TRP) architecture using a Long Short-Term Memory (LSTM) and Extreme Gradient Boosting (XGBoost) model to improve both travel efficiency and environmental sustainability in route selection. This model incorporates carbon emissions directly into the route planning process by unifying user preferences, location recommendations, route optimization, and multimodal vehicle selection within a comprehensive framework. By merging environmental sustainability with user-focused travel planning, it generates personalized, practical, and low-carbon travel routes. The carbon emissions observed with TRP’s artificial intelligence (AI) recommendation route are presented comparatively with those of the user-determined route. XGBoost, Random Forest (RF), Categorical Boosting (CatBoost), Light Gradient Boosting Machine (LightGBM), (Extra Trees Regressor) ETR, and Multi-Layer Perception (MLP) models are applied to the TRP model. LSTM is compared with Recurrent Neural Networks (RNNs) and Gated Recurrent Unit (GRU) models. Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE), and Normalized Root Mean Square Error (NRMSE) error measurements of these models are carried out, and the best result is obtained using XGBoost and LSTM. TRP enhances environmental responsibility awareness within travel planning by integrating sustainability-oriented parameters into the decision-making process. Unlike conventional reservation systems, this model encourages individuals and organizations to prioritize eco-friendly options by considering not only financial factors but also environmental and socio-cultural impacts. By promoting responsible travel behaviors and supporting the adoption of sustainable tourism practices, the proposed approach contributes significantly to the broader dissemination of environmentally conscious travel choices. Full article

Fishery resources of tuna serve as a vital source of global protein. This study investigates the key environmental drivers influencing the spatial distribution of yellowfin tuna (Thunnus albacares) in the western tropical Pacific Ocean. A comprehensive dataset was constructed by linking the catch per unit effort (CPUE) from 43 Chinese longline fishing vessels (2008–2019) with 24 multi-source environmental variables. To accurately model this complex relationship, a total of 16 machine learning regression models, including advanced ensemble methods like Light Gradient Boosting Machine (LightGBM), Random Forest, and Categorical Boosting Regressor (CatBoost), were evaluated and compared using multiple performance metrics (e.g., Coefficient of Determination [R²], Root Mean Squared Error [RMSE]). The results indicated that the Light Gradient Boosting Machine (LightGBM) model achieved superior performance, demonstrating excellent nonlinear fitting capabilities and generalization ability. For robust feature interpretation, the study employed both the model’s internal feature importance metrics and the SHapley Additive exPlanations (SHAP) method. Both approaches yielded highly consistent results, identifying temporal (month), spatial (longitude, latitude), and key seawater temperature indicators at intermediate depths (T450, T300, T150) as the most critical predictors. This highlights significant spatiotemporal heterogeneity in the distribution of Thunnus albacares. The analysis suggests that mid-layer ocean temperatures directly influence catch rates by governing the species’ vertical and horizontal movements. In contrast, large-scale climate indices such as the Oceanic Niño Index (ONI) exert indirect effects by modulating ocean thermal structures. This research confirms the dominance of spatiotemporal and thermal variables in predicting yellowfin tuna distribution and provides a reliable, data-driven framework for supporting sustainable fishery management, resource assessment, and operational forecasting. Full article

Low back pain (LBP) is one of the leading causes of disability in the world's population, yet there are limitations in providing an objective clinical assessment due to its widespread nature. In this work, five machine learning models (LightGBM, XGBoost, HistGradientBoosting, GradientBoosting, and StackingRegressor) were compared to predict trunk mobility based on inertial sensor data. There were 77 individuals with a total of 2160 movement samples of flexion–extension, rotation, and lateralization. Synthetic data augmentation and normalization were performed to be able to work with the data efficiently. Mean absolute error (MAE), mean square error (MSE), and R2 were used to evaluate model performance. Additionally, ANOVA and Tukey’s HSD were used to assess the statistical significance of the models. GradientBoostingRegressor was found to produce the lowest error and statistical significance in flexion–extension and lateralization, while StackingRegressor produced the best error in rotation. The results indicate that inertial sensors and machine learning (ML) can be applied to predict mobility, facilitating personalized rehabilitation and reducing costs. The present study demonstrates that predictive trunk motion modeling can facilitate clinical monitoring and help reduce socioeconomic limitations in patients. Full article

Accurate estimation of lithium-ion battery (LIB) state of health (SOH) underpins safe operation, predictive maintenance, and lifetime-aware energy management. Despite recent advances in machine learning (ML), systematic benchmarking across heterogeneous real-world cells remains limited, often confounded by data leakage and inconsistent validation. Here, we establish a leakage-averse, cross-battery evaluation framework encompassing 32 commercial LIBs (B5–B56) spanning diverse cycling histories and temperatures (≈4 °C, 24 °C, 43 °C). Models ranging from classical regressors to ensemble trees and deep sequence architectures were assessed under blocked 5-fold GroupKFold splits using RMSE, MAE, R² with confidence intervals, and inference latency. The results reveal distinct stratification among model families. Sequence-based architectures—CNN–LSTM, GRU, and LSTM—consistently achieved the highest accuracy (mean RMSE ≈ 0.006; per-cell R² up to 0.996), demonstrating strong generalization across regimes. Gradient-boosted ensembles such as LightGBM and CatBoost delivered competitive mid-tier accuracy (RMSE ≈ 0.012–0.015) yet unrivaled computational efficiency (≈0.001–0.003 ms), confirming their suitability for embedded applications. Transformer-based hybrids underperformed, while approximately one-third of cells exhibited elevated errors linked to noise or regime shifts, underscoring the necessity of rigorous evaluation design. Collectively, these findings establish clear deployment guidelines: CNN–LSTM and GRU are recommended where robustness and accuracy are paramount (cloud and edge analytics), while LightGBM and CatBoost offer optimal latency–efficiency trade-offs for embedded controllers. Beyond model choice, the study highlights data curation and leakage-averse validation as critical enablers for transferable and reliable SOH estimation. This benchmarking framework provides a robust foundation for future integration of ML models into real-world battery management systems. Full article

Computing the temporal variation in clearwater scour depth around abutments is important for bridge foundation design. To reach the equilibrium scour depth at bridge abutments takes a very long time. However, the corresponding times under prototype conditions can yield values significantly greater than the time to reach the design flood peak. Therefore, estimating the temporal variation in scour depth is necessary. This study evaluates multiple machine learning (ML) models to identify the most accurate method for predicting scour depth (Ds) over time using experimental data. The dataset of 3275 records, including flow depth (Y), abutment length (L), channel width (B), velocity (V), time (t), sediment size (d₅₀), and Ds, was used to train and test Linear Regression (LR), Random Forest Regressor (RFR), Support Vector Regression (SVR), Gradient Boosting (GBR), XGBoost, LightGBM, and KNN models. Results demonstrated the superior performance of AI-based models over conventional regression. The RFR model achieved the highest accuracy (R² = 0.9956, Accuracy = 99.73%), followed by KNN and GBR. In contrast, the conventional LR model performed poorly (R² = 0.4547, Accuracy = 57.39%). This study confirms the significant potential of ML, particularly ensemble methods, to provide highly reliable scour predictions, offering a robust tool for enhancing bridge design and safety. Full article