Search Results (13,782)

Background: Accurate preoperative prediction of length of hospital stay (LOS) after surgery for colorectal liver metastases (CRLMs) could improve patient counselling and resource planning, yet reliable risk tools are lacking. We aimed to develop an interpretable machine learning model predicting LOS following first-time liver-directed surgery for CRLMs. Methods: In this multicenter cohort study, we included patients who underwent first-time liver resection, ablation, or a combination for CRLMs at three Danish hepatobiliary centers between 2016 and 2023. Preoperative features from two national registries were used to train Elastic Net, Random Forest, HistGradientBoosting, and Explainable Boosting Machine (EBM) algorithms. Hyperparameters were optimized using five-fold cross-validation. Performance was evaluated on a 20% hold-out test sample using mean absolute error (MAE) with bootstrapped 95% confidence intervals (CIs). Results: Among 915 patients, median LOS was 4.0 days (interquartile range (IQR) 3.0–6.0). All four algorithms achieved comparable prediction error (MAE 3.0–3.1 days). The EBM (MAE 3.1 days, 95% CI 2.6–4.3) algorithm was selected for its inherent interpretability. Surgical approach was the strongest predictor, where percutaneous and laparoscopic approaches were associated with reductions of 1.9 and 1.2 days, respectively. Tumor burden, including number of lesions and largest lesion diameter, showed progressive non-linear associations with longer stays. Nonetheless, overall explained variance was low (R² ≤ 0.10), and calibration showed systematic underestimation of stays beyond five days. Conclusions: An inherently interpretable machine learning model matched the predictive performance of opaque algorithms for LOS after CRLM surgery, although overall predictive accuracy was modest and longer stays were underestimated. Explainability analysis identified surgical approach and tumor burden as the most influential predictors. External validation in healthcare systems with different discharge practices is warranted. Full article

Efficient management of soil–water resources is critical for global food security under intensifying climatic and demographic pressures. This review provides a comprehensive synthesis of artificial intelligence (AI) and distributed deep learning methodologies applied to soil–water interactions in precision agriculture. The physical and hydraulic foundations of soil–water systems—including water retention, unsaturated flow governed by the Richards equation, and soil degradation processes—are examined and situated within a unified framework of AI-based modeling and decision support. Classical machine learning (ML) algorithms (Random Forests, Support Vector Machines, gradient boosting) and deep learning architectures (convolutional neural networks, long short-term memory networks, transformers) are evaluated with respect to their capacity to predict soil moisture dynamics, estimate hydraulic properties, support smart irrigation scheduling, and generate digital soil maps at field-to-regional scales. Distributed training paradigms, federated learning for privacy-preserving multi-farm analytics, and edge AI deployment on low-power IoT hardware are assessed as enabling infrastructures for scalable agricultural intelligence. This review further addresses explainability, uncertainty quantification, and ethical dimensions inherent to AI-driven agricultural systems. Key challenges—including training data scarcity in data-poor regions, model interpretability, integration with physics-based hydrological models, and real-time deployment constraints—are critically discussed. Prospective research directions encompass physics-informed neural networks, foundation models for earth observation, autonomous digital twins of soil–water systems, and federated learning architectures aligned with data sovereignty frameworks. The synthesis underscores AI’s transformative potential for sustainable agricultural water management while delineating the technical and sociotechnical barriers that must be resolved to realize this potential at a global scale. Full article

Power cycling reliability is one of the most widely used frameworks to evaluate the lifetimes of power semiconductor switching devices from a thermal stress perspective. Experimental tests can be used to predict their lifetimes under operating conditions. An estimation of the number of cycles to failure $N_f$ can also be given by several lifetime models, which express the number of cycles to end of life as a function of empirical coefficients. In the existing literature, these empirical coefficients are generally estimated using the classical least squares method (to find the best-fitting line through data points), where outliers are removed using the Random Sample Consensus algorithm. The aim of this paper is to present a general strategy for the calculation of empirical coefficients for different lifetime models, such as Coffin–Manson, Coffin–Manson–Arrhenius, Norris–Landzberg, and simplified Bayerer, aiming at minimizing the number of required experimental tests. The results show that the number of experimental trials required varies between two and four, depending on the number of empirical coefficients to be determined, which is specific to the lifetime model used. Furthermore, a limited number of experimental data points are selected to avoid any degradation in accuracy. The accuracy of coefficient estimation is significantly improved by excluding outliers: some relative errors decrease by 25%. Additionally, each empirical coefficient is determined under specific thermal stress conditions, such as a constant junction temperature swing $\mathsf{\Delta }{T}_j$, constant current per bond wire I, constant cycling frequency f, or constant mean junction temperature $T_{jm}$. Furthermore, a limited number of experimental data are selected to avoid any degradation in accuracy due to outliers. Moreover, this general method can be applied to all power devices, such as IGBTs or MOSFETs. Finally, the limitations of the analytical solution for the Scheuermann lifetime model are discussed. Full article

The Al-Ahsa Oasis region is an important agricultural area; however, continuous spatial–temporal monitoring is essential to assess and mitigate the impacts of climate change and land use change. The current study examines environmental and land cover changes in the Al-Ahsa Oasis region from 1990 to 2025 by utilizing spectral indices derived from multiple satellites. Multi-temporal Landsat imagery (Landsat 5, 8, and 9) was processed in Google Earth Engine (GEE) to derive key biophysical indicators, including the Normalized Difference Vegetation Index (NDVI), land surface temperature (LST), and bare soil index (BSI). Supervised classification techniques were employed to generate LULC maps for each time step, enabling the assessment of spatiotemporal land cover dynamics. In addition, a random forest (RF) machine learning algorithm was applied to accurately quantify and map the distribution of palm trees across the study area. The results showed that NDVI values fluctuated between −0.19 and 0.75 during the period from 1990 to 2025. Higher vegetation density was observed in central and eastern areas, with maximum values of −0.44–0.75 in 2025. The higher LST was observed in 2025, with a range of 34.7 to 54.6 °C, and the lower LST was observed in 1990 with a range 28.7 to 48.34 °C. BSI values decreased from −0.40 to 0.46 between 1990 and 2025 to a more variable range of −0.27 to 0.36, indicating reduced soil exposure. The classification of LULC numerical data shows a rapid rise in urban development of 67.19% and a 25% decrease in vegetation area. Furthermore, the results of the RF model indicate that palm tree area increased by 16.23% from 1990 to 2025, with overall accuracy of 98.15, and kappa coefficient of 0.962. This research highlights that urban expansion impacts environmental indicators such as LST, while the increasing trend of NDVI could support the palm trees expansion. This study finds valuable information for policymakers and land use planners to develop sustainable urban growth strategies, protect agricultural lands, and enhance oasis ecosystem resilience. Combined remote-sensing-based monitoring into regional planning frameworks can inform decision making for balancing urban development, environmental protection, and long-term agricultural sustainability in the Al-Ahsa Oasis. Full article

The Graduate Student Evolutionary Algorithm (GSEA) has demonstrated promising optimization capability in several engineering tasks; however, its performance may deteriorate when dealing with high-dimensional and complex multimodal problems due to insufficient adaptive search behavior, weak diversity preservation, and stagnation during later optimization stages. To alleviate these limitations, this paper proposes a Multi-Strategy Improved Graduate Student Evolutionary Algorithm (MIGSEA) for numerical optimization and artistic image multi-threshold segmentation. First, an adaptive mentor-guided learning mechanism is introduced to dynamically regulate the influence of mentors and peers throughout the optimization process, enabling a more effective transition from global exploration to local exploitation. Second, an elite–random cooperative learning strategy is designed to combine high-quality solution guidance with stochastic perturbation, thereby improving population diversity and enhancing the ability to escape local optima. Third, a stagnation-aware local refinement mechanism is developed to activate adaptive neighborhood search when the optimization process becomes trapped, which further accelerates convergence and improves solution precision. To verify the effectiveness of the proposed algorithm, MIGSEA is evaluated on the IEEE CEC2017 and CEC2020 benchmark suites and compared with 11 advanced metaheuristic algorithms under identical experimental conditions. Experimental results demonstrate that MIGSEA achieves competitive optimization accuracy, convergence speed, robustness, and statistical superiority in most benchmark functions. Furthermore, MIGSEA is applied to Otsu-based artistic image multi-threshold segmentation using multiple benchmark images with different threshold levels. Quantitative evaluation based on PSNR, FSIM, and SSIM, together with visual analysis, confirms that the proposed method can generate more accurate and visually consistent segmentation results than existing competitors. Overall, the proposed MIGSEA provides an effective and robust optimization framework for both benchmark optimization and practical image segmentation applications. Full article

Flow-cell ultrasonication of gelatinized rice flour slurries alters cultivar-dependent water solubility, viscosity, and retrogradation of pregelatinized rice flour, properties important for plant-based beverages and convenience foods. We tested whether cultivar-level composition descriptors, amylose, protein, and fiber, can represent cultivar-associated variation in ultrasonication responses while separating process-only prediction, within-domain cultivar representation, and unseen-cultivar transfer. Six rice cultivars were processed across nine amplitude-time combinations and two slurry concentrations. Water solubility index, apparent viscosity at a shear rate of 50 s⁻¹, and setback viscosity were modeled using ElasticNet, partial least squares regression, support vector regression, random forest, and extreme gradient boosting. Three input formulations were compared: process variables alone, process variables plus composition descriptors, and process variables plus cultivar identity. Repeated nested group cross-validation showed insufficient process-only prediction and substantial improvement from composition descriptors. Within-domain validation showed comparable composition-descriptor and cultivar-identity performance under nonlinear algorithms. However, because cultivar identity is undefined for absent cultivars, leave-one-cultivar-out transfer of the composition-descriptor model remained uncertain. Cross-fitted Shapley additive explanations showed predictions used process and composition variables. For the validated cultivar-process domain, this approach can screen cultivar-process combinations for beverage and convenience-food applications, but replacing categorical source identifiers with continuous descriptors requires explicit transfer validation. Full article

With the advancement of energy market reform, integrated energy systems (IESs) have achieved rapid development. Considering insufficient research on an electricity–heat coupled master–slave game and the local optimum defect of traditional algorithms, this paper proposes a Stackelberg game optimization strategy for IES considering integrated demand response (IDR), with microgrid operator (MGO) as the leader and load aggregator (LA) as the follower. Firstly, an IDR model containing rigid, shiftable electric loads and reducible thermal loads is established, and a bi-level game model is built: the upper MGO optimizes electricity and heat pricing to maximize profit, while the lower LA adjusts flexible loads for maximum consumer surplus. Secondly, an improved snake optimizer (ISO) is constructed via Hammersley sequence initialization, Lévy flight and random perturbation and combined with quadratic programming (QP) to form the ISO-QP hybrid solving method. Benchmark function and CEC2017 tests verify the superior convergence and stability of ISO against multiple classical intelligent algorithms. Case simulation obtains the Stackelberg equilibrium result, and repeated experiments and parameter sensitivity analysis verify model robustness. Results show that the proposed method smooths load fluctuations via price guidance and synchronously improves MGO revenue and LA consumer surplus on the premise of guaranteed user satisfaction. Full article

With the deep integration of artificial intelligence and big data, intelligent optimization algorithms have become key tools for solving many complex problems. However, as problem scale and complexity grow rapidly, the performance of traditional algorithms often faces significant challenges. The Teaching Learning Based Optimization algorithm has attracted widespread attention for its simple structure, few parameters, and high solution efficiency, and has been successfully applied across various engineering and scientific fields. Nevertheless, when dealing with high-dimensional, multimodal global optimization problems and real-world applications, the standard Teaching Learning Based Optimization still exhibits certain limitations, such as reduced accuracy of the optimal solution due to insufficient initial population diversity, and difficulty in escaping local optima caused by premature convergence. To address these issues, this paper proposes an Improved Teaching Learning Based Optimization algorithm. The improved ITLBO upgrades original TLBO from three perspectives: first, a population interaction strategy combining chaotic disturbance and Gaussian mutation is designed to enrich initial population diversity; second, bipolar cooperative search utilizing dynamic weighting of optimal and worst individuals balances global exploration and local exploitation to avoid premature convergence; third, oscillatory random mapping learning with sinusoidal oscillation factor periodically perturbs individuals to continuously replenish population diversity in iterations. Numerical results show that the proposed method exhibits superior convergence performance and stability on classical global optimization benchmarks. Furthermore, the algorithm is applied to practical cloud resource scheduling problems, and experimental outcomes verify that ITLBO improves solution accuracy by approximately one order of magnitude over original TLBO and reduces small-scale cloud scheduling cost by 12% while achieving preferable robustness. Full article

PM_2.5 is a major air pollutant that threatens urban air quality and public health. Its concentration is influenced by both meteorological conditions and air pollutants, exhibiting complex nonlinear and temporal characteristics. Traditional statistical methods are limited in their ability to model complex relationships among environmental variables, while machine learning models still require improvements in hyperparameter optimization and interpretability. Therefore, developing an accurate and interpretable PM_2.5 estimation model remains an important research objective. This study used daily air-quality and meteorological data collected in Wuhan from 2016 to 2025 to develop six machine learning models: Decision Tree (DT), Random Forest (RF), XGBoost, LightGBM, Support Vector Machine (SVM), and Multilayer Perceptron (MLP). The Particle Swarm Optimization (PSO) algorithm was employed to optimize the hyperparameters of these models. By comparing the root mean square error (RMSE), coefficient of determination (R²), and mean absolute error (MAE) of each model on both the training and test sets, the PSO-MLP model was identified as the best-performing model. Furthermore, the Shapley Additive Explanations (SHAP) method was applied to perform both global and local interpretation analyses of the best-performing model. The results indicate that the PSO-MLP model achieved the highest estimation performance among all evaluated models, with an R² value of 0.746 on the test set. SHAP analysis revealed that CO, Temperature (Temp), and NO₂ were the most influential predictors, while all variables exhibited distinct nonlinear relationships with PM_2.5 concentration. These findings may contribute to PM_2.5 concentration estimation, air-quality management, and environmental decision-making. Full article

With the adoption of more ambitious emission reduction strategies in the shipping industry by the International Maritime Organization and the resulting stricter greenhouse gas emission reduction requirements, it is particularly important for all stakeholders in the global maritime shipping industry to assess the energy efficiency of shipping vessels. Forming predictive capabilities for ship fuel consumption and Carbon Intensity Indicator (CII) annual ratings, for example, are two important works. This article adopted 14 different algorithms in three categories of data-driven approaches, i.e., statistics, machine learning and deep learning, including polynomial regression, ridge regression, adaptive boosting, categorical boosting, elastic net, etc., and built the ship fuel consumption prediction model using ship noon report as the data source. The prediction accuracy and computational efficiency of model training were compared based on metrics of coefficient of determination, mean absolute percentage error and floating-point operations per amount of training data. Cross-validations were performed for all 14 algorithms to analyze their sensitivities to their respective tuned parameters. Comparisons indicated that algorithms of the statistics approach were sensitive to the quality of the data source, compared with the machine learning and the deep learning approaches. The accuracy of the elastic net algorithm was sensitive to the tuned parameters. Two algorithms, light gradient boosting machine and random forest, were selected based on their performances of prediction accuracy and computational efficiency of model training. Then, the selected algorithms were separately combined with long short-term memory as the time-series prediction algorithm to form their respective coupled framework. Both of the coupled frameworks achieved successful prediction of the CII annual discriminant and rating of the studied ships. The prediction accuracy was validated to be sufficient. Full article

Optimizing the injection molding of short-fiber-reinforced thermoplastics (SFRTs) is a persistent challenge due to the complex interplay between processing parameters and final mechanical performance. To address this, we developed and validated a machine learning (ML) pipeline to maximize both the tensile strength and Charpy impact resistance in polyamide 6 with 30% glass fiber (PA6-GF30). Through a designed experimental campaign, we systematically varied four key process parameters—melt temperature (260–300 °C), injection pressure (600–1000 bar), packing pressure (400–800 bar), and cooling time (15–35 s). The resulting dataset was used to train and compare three different regression models: Random Forest (RF), Gradient Boosting (GB), and Support Vector Regression (SVR). Our findings indicate that the Gradient Boosting (GB) algorithm yielded the most reliable predictions, significantly outperforming the other evaluated models. Further analysis using SHAP (Shapley Additive exPlanations) identified packing pressure as the dominant factor influencing tensile strength (contributing approximately 40% to the prediction), while melt temperature emerged as the key driver for impact resistance (around 35% contribution). By integrating our best-performing GB model with a multi-objective genetic algorithm, we identified an optimal set of parameters that simultaneously enhances both mechanical properties. Among the evaluated models (Random Forest, Support Vector Regression, and Gradient Boosting), the Gradient Boosting algorithm achieved the highest predictive accuracy. Compared to the baseline condition (280 °C melt temperature, 800 bar injection pressure, 600 bar packing pressure, 25 s cooling time), experimental validation of these optimized settings demonstrated substantial improvement: tensile strength increased from 145 MPa to 171 MPa (an 18% enhancement), and impact resistance rose from 45 kJ/m² to 55 kJ/m² (a 22% gain). This work establishes that an integrated ML and optimization framework can serve as a transformative approach for high-precision manufacturing of advanced engineering polymers. The primary novelty of this work lies in the development of a fully integrated, bias-free methodological framework that explicitly couples physical interpretability with multi-objective optimization, bridging the critical gap between black-box predictions and actionable industrial insights. Full article

Complex optimization problems are widely encountered in engineering design, intelligent manufacturing, communication systems, and wireless sensor network deployment. However, the original Enterprise Development Optimization Algorithm (EDOA) still suffers from insufficient population diversity, weak search guidance, and limited adaptability in balancing exploration and exploitation when solving high-dimensional and multimodal optimization problems. To address these issues, this paper proposes a Multi-Strategy Improved Enterprise Development Optimization Algorithm (MIEDOA). First, a Strategic Diversification Initialization (SDI) strategy is developed by integrating Sobol sequence sampling, random initialization, and Gaussian perturbation to improve the diversity and distribution quality of the initial population. Second, an Organizational Synergy Learning (OSL) mechanism is introduced to enhance search guidance through the collaborative utilization of elite information, population mean information, and peer interaction. Third, an Adaptive Governance with Feedback Regulation (AGFR) strategy is designed to dynamically regulate the exploration–exploitation behavior according to the current population fitness state. The proposed MIEDOA is evaluated on the CEC2017 and CEC2020 benchmark suites and compared with representative EDOA variants, CEC winner algorithms, and other advanced optimization methods. The experimental results indicate that MIEDOA generally achieves competitive performance in terms of solution quality, convergence behavior, and robustness across different benchmark scenarios. In addition, strategy effectiveness analysis, parameter sensitivity analysis, and statistical tests further provide evidence supporting the effectiveness of the proposed strategies. Finally, MIEDOA is applied to a three-dimensional wireless sensor network deployment problem. The results suggest that the proposed algorithm can obtain competitive deployment solutions and satisfactory coverage performance under different node scales, demonstrating its potential applicability to practical engineering optimization problems. Full article

School buses serve as the primary mode of transportation for children traveling to and from school, and their safety measures represent a critical safeguard for children’s lives. Nevertheless, incidents in which children are left unattended on school buses—due to inadequate supervision or the children’s own actions—occur with notable frequency and can lead to fatal outcomes. To mitigate or prevent such tragedies, this paper proposes an in-vehicle thermal imaging solution based on infrared array sensors, integrated with a dedicated algorithm to detect whether a child has been left behind in the school bus. The system collects background temperature, presence temperature, and real-time temperature data inside the bus using infrared array sensors. By comparing the real-time temperature difference against a predefined presence temperature difference threshold, the algorithm determines whether a child is present under the current thermal conditions. It then verifies whether the number of positive detections within a specified temperature range meets a preset presence count threshold, thereby reaching a final decision regarding child presence. Experiments identified optimal parameters: a temperature range of 26–33 °C, a double-difference threshold (ε = 1), and a presence count threshold (P = 4). Random testing demonstrated that the proposed technical solution and algorithm achieve an overall detection success rate of 92.5%. This study develops a low-cost, easily deployable, non-contact thermal imaging method capable of identifying forgotten children on school buses with satisfactory accuracy. By detecting retention before harm occurs, the approach enhances the safety of children traveling by school bus. Full article

This article presents a method for calibration of dynamic vehicle weighing systems (WIM—Weigh-In-Motion) involving the calibration of all WIM stations operating within a given road network segment as a single process. A key assumption of the method is the presence of at least one scale with significantly higher accuracy than the calibrated systems in this part of road network. This reference scale function may be played by a static scale, slow-pass scale (LS-WIM—Low-Speed WIM) for measurement of vehicle axle load or by a selected WIM system with heightened accuracy. Both the reference scale and all systems undergoing calibration must be equipped with a system for the automatic recognition of vehicle registration number plates. The reference scale makes it possible to determine axle load values considered as benchmark values. Then, for each vehicle weighed on the reference scale and subsequently on any WIM system operating within the analysed area, the relative difference between the reference result and the WIM system measurement is calculated with respect to the reference value. This difference forms the basis for the operation of the algorithm estimating the coefficients of the static characteristic of the calibrated WIM system (so-called calibration coefficients), which are then used to determine corrected weighing results. The estimation of the coefficients is updated after each identified vehicle that has previously been weighed on the reference scale is considered. The article presents both the results of simulations and experimental studies concerning the proposed spatial method of calibration. The results obtained allow for an assessment of the effectiveness of the proposed solution. As can be seen from the analyses conducted, this method leads to a significant reduction in systematic error of vehicle weight measurement. Unfortunately, it does not eliminate random errors. The spatial calibration approach described in this paper has certain limitations. The main ones include the impact of ANPR system errors on calibration effectiveness, cases where a vehicle is unloaded or loaded between WIM stations, and the propagation of systematic errors from the reference systems to the other WIM systems. A significant advantage of the proposed spatial calibration method is that it can operate effectively using weighing data from a single reference WIM system and does not require heavy traffic volumes. Full article

Non-contact magnetic sensing technology is widely adopted in transmission line online monitoring scenarios including current measurement and fault location for its non-contact measurement capability, strong environmental robustness and low deployment cost. However, existing magnetic-sensing-based galloping monitoring methods suffer from two critical limitations: no theoretical guidance is provided for sensor placement, and a high false detection rate is observed under current fluctuation conditions. To address these issues, a novel transmission line galloping monitoring method based on spatial magnetic field distribution features is proposed in this paper. A conductor galloping-power frequency magnetic field coupling model is first established to derive the optimal magnetic sensor array arrangement strategy. Subsequently, a galloping detection algorithm fusing multi-node frequency-domain features and phase difference information is proposed to eliminate current fluctuation induced false detection. Simulations conducted based on actual 500 kV transmission line parameters and verification tests carried out on a scaled-down laboratory platform confirm that reliable galloping detection can be realized by the proposed method under both current low-frequency oscillation and random fluctuation scenarios. With advantages of non-contact deployment, high anti-interference performance and detection accuracy, the proposed method has promising application potential in engineering-oriented high-voltage transmission line monitoring. Full article