Gradient Boosting Regression Research Articles

Rapid quantitative and qualitative analysis of talcum powder in wheat flour by microwave detection combined with multivariate analysis

Rapid and accurate monitoring of talcum powder in wheat flour is of great practical importance for maintaining market order, public health and food safety. In this study, the potential of a home-made microwave detection system for rapid quantitative and qualitative detection of talcum powder adulteration in wheat flour at 2.5–11.5 GHz was investigated. Useful microwave spectral information is selected using three feature selection strategies, the bootstrapping soft shrinkage (BOSS), the multiple feature spaces ensemble with least absolute shrinkage and selection operator (MEF-LASSO), and the competitive adaptive reweighted sampling (CARS). The eXtreme Gradient Boosting (XGBoost) classification and regression models are constructed and evaluated, and three nature-inspired optimization algorithms (NIOA) are used for parameter optimization: the Osprey Optimization Algorithm (OOA), the Crested Porcupine Optimizer (CPO), and the Sparrow Search Algorithm (SSA). The experimental results show that both feature selection strategies and NIOA algorithms can effectively improve the model performance. The SSA-XGBoost classification model has the highest prediction accuracy, and its prediction accuracy and F1-score can reach 100 % and 1, respectively. The MEF-LASSO-SSA-XGBoost regression model has strong robustness. The root mean square error (RMSE) was 1.76 %, the prediction coefficient of determination (R2) was 0.98, and the ratio of performance to deviation (RPD) was 8.66. The results of this study showed that the combination of microwave technology and machine learning model with multivariate analysis method can realize the rapid and accurate determination of qualitative and quantitative talcum powder in wheat flour.

Study on the design method of multi-component industrial solid waste low carbon cementitious material with cement as the activator

The relationship between microstructure and mechanical properties of multi-component solid waste low-carbon cementitious materials has been widely pay attention to. However, industrial solid waste is a complex multi-component system with many variable factors, which makes it difficult to design the formulation of cementitious materials. This paper pioneered the application of machine learning (ML) models, algorithms and error rates to analyze the compressive and flexural strength of fly ash-based pastes. Coefficient of determination (R2), mean squared error (MSE), root mean square error (RMSE), mean absolute error (MAE) and a20-index were used to evaluate robustness. X-ray diffraction (XRD), scanning electron microscope (SEM) and Brunauer-Emmett-Taylor (BET) were carried out to analyze evolution of cementitious materials. The evaluation results of ML models exhibited that the Gradient boosting regression (GBR) model had the best determination parameters and a steep normal distribution fitting curve with an a20-index of 0.861. GBR model exhibited the best robustness. The key factors of fly ash-based cementitious materials were identified by Pearson's coefficient, which was benefit to determine the formulation of multi-component solid waste low-carbon cementitious materials. Furthermore, experiments also demonstrated that the optimum ratio of multi-component solid waste low carbon cementitious material was 10 % gypsum, 10 % metakaolin, 45 % fly ash, 15 % slag and 20 % cement, respectively. It was worth noting that the compressive strength of this kind of multi-component solid waste low-carbon cementitious materials reached 35 MPa, which was superior to the mechanical properties of P·O 32.5 cement. The results of phase, SEM images and pore structure distribution showed that the synergistic effect of the multi-component solid waste materials effectively filled the material voids and also facilitated the formation of a variety of gelatinous materials through gelling reactions in the late stage (14–28 d). This work will promote the resource utilization of industrial solid waste, contribute to carbon reduction, and can accelerate the green revolution of concrete.

Predicting sunspot number from topological features in spectral images I: Machine learning approach

This study presents an advanced machine learning approach to predict the number of sunspots using a comprehensive dataset derived from solar images provided by the Solar and Heliospheric Observatory (SOHO). The dataset encompasses various spectral bands, capturing the complex dynamics of solar activity and facilitating interdisciplinary analyses with other solar phenomena. We employed five machine learning models: Random Forest Regressor, Gradient Boosting Regressor, Extra Trees Regressor, Ada Boost Regressor, and Hist Gradient Boosting Regressor, to predict sunspot numbers. These models utilized four key heliospheric variables — Proton Density, Temperature, Bulk Flow Speed and Interplanetary Magnetic Field (IMF) — alongside 14 newly introduced topological variables. These topological features were extracted from solar images using different filters, including HMIIGR, HMIMAG, EIT171, EIT195, EIT284, and EIT304. In total, 60 models were constructed, both incorporating and excluding the topological variables. Our analysis reveals that models incorporating the topological variables achieved significantly higher accuracy, with the r2-score improving from approximately 0.30 to 0.93 on average. The Extra Trees Regressor (ET) emerged as the best-performing model, demonstrating superior predictive capabilities across all datasets. These results underscore the potential of combining machine learning models with additional topological features from spectral analysis, offering deeper insights into the complex dynamics of solar activity and enhancing the precision of sunspot number predictions. This approach provides a novel methodology for improving space weather forecasting and contributes to a more comprehensive understanding of solar-terrestrial interactions.

Within-canopy variation in chlorophyll fluorescence parameters can be well captured by vertical patterns of leaf biophysical and chemical traits.

The strong correlation between chlorophyll fluorescence (ChlF) parameters and photosynthesis highlights the need for a comprehensive spatial representation of ChlF parameters within the canopy. Such an approach is essential to advance our understanding and to improve the representation and modeling of water and carbon fluxes at scales ranging from the leaf to the canopy level. However, the challenge remains in determining how to effectively describe and track the variability of ChlF parameters within the canopy. In this study, we determined the variation in leaf biophysical and chemical traits and ChlF parameters along the vertical height of the canopy for several species in a temperate deciduous forest. We observed general associations of height with leaf biophysical and chemical traits and ChlF parameters, although these relationships were species-dependent. In addition, leaf biophysical and chemical traits, particularly light-harvesting pigments, showed significant effects on ChlF parameters. To effectively track variation in ChlF parameters within the canopy, we used gradient-boosted regression (GBR) models driven by leaf traits and species, which explained more than 80% of the variation in all ChlF parameters. Our study demonstrates the feasibility of utilizing leaf biophysical and chemical traits to predict vertical variation in ChlF parameters and provide supportive data for modeling canopy photosynthesis.

A Comparative Analysis of Machine Learning Approaches for Evaluating the Compressive Strength of Pozzolanic Concrete

This study leverages machine learning techniques to predict pozzolanic concrete's compressive strength accurately. Using artificial neural networks (ANN), random forest (RF), and gradient boosting regressor (GBR) models trained on a dataset of 482 samples, the study divides the data into 70% training and 30% testing subsets with seven input parameters. Model performance is assessed through metrics like coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE). The RF model excels, achieving R2 values of 0.976 in training and 0.964 in testing, along with the lowest RMSE (2.84 MPa) and MAE (2.05 MPa) during training and RMSE values of 7.81 MPa and MAE values of 5.89 MPa during testing, demonstrating superior predictive accuracy. Sensitivity analysis highlights the pivotal role of cement as an input parameter, contributing significantly to the model's accuracy. Employing K-fold cross-validation confirms the RF model's robustness with an average R2 value of 0.959. This research underscores the RF model's reliability and effectiveness in forecasting pozzolanic concrete compressive strength, with practical applications for concrete optimization and construction practices, establishing it as the preferred choice compared to other machine learning models. IUBAT Review—A Multidisciplinary Academic Journal, 7(1): 90-122

Evaluation of empirical and machine learning models for predicting shear wave velocity of granular soils based on laboratory element tests

Shear wave velocity (Vs) is crucial for designing geotechnical systems subjected to dynamic loads, especially in seismically active regions. The shear wave velocity of geomaterials can be determined using in situ and laboratory tests. However, due to time and cost limitations, the Vs is not easily available in most projects. Various empirical models have been developed by researchers for predicting the shear wave velocity of geomaterials. However, most of these models have been developed for specific soil types and loading characteristics. In this work, for predicting the shear wave velocity of granular soils using various combinations of input parameters, various empirical models were proposed. Furthermore, machine learning (ML) methods were utilized to predict the Vs. The suggested models consider the impact of grading characteristics such as fine content (FC), gravel content (GC), median particle size (D50), uniformity coefficient (Cu), and coefficient of curvature (Cc), as well as void ratio (e), mean effective confining pressure (σm′), consolidation stress ratio (KC), and specimen preparation techniques for reconstitution of specimens. To achieve this, a series of bender element tests were performed on various sand and gravel mixtures. Furthermore, data from previous studies were also used. So, the study utilized 513 data points from laboratory element experiments conducted on granular soils. For predicting the Vs of granular soils, four empirical models and three ML models, namely Artificial Neural Network (ANN), Support Vector Regression (SVR), and Gradient Boosting Regression (GBR), were developed in this study. The findings showed that the ANN model outperforms the other comparative models in terms of accuracy and error. While the empirical models may serve as useful tools for initial Vs estimation in construction projects, the study primarily highlights the significance of using ML methods to enhance the prediction accuracy of Vs based on the available soil properties.

Meta-heuristic optimization algorithms based feature selection for joint moment prediction of sit-to-stand movement using machine learning algorithms

The sit-to-stand (STS) movement is fundamental in daily activities, involving coordinated motion of the lower extremities and trunk, which leads to the generation of joint moments based on joint angles and limb properties. Traditional methods for determining joint moments often involve sensors or complex mathematical approaches, posing limitations in terms of movement restrictions or expertise requirements. Machine learning (ML) algorithms have emerged as promising tools for joint moment estimation, but the challenge lies in efficiently selecting relevant features from diverse datasets, especially in clinical research settings. This study aims to address this challenge by leveraging metaheuristic optimization algorithms to predict joint moments during STS using minimal input data. Motion analysis data from 20 participants with varied mass and inertia properties are utilized, and joint angles are computed alongside simulations of joint moments. Feature selection is performed using the Manta Ray Foraging Optimization (MRFO), Marine Predators Algorithm (MPA), and Equilibrium Optimizer (EO) algorithms. Subsequently, Decision Tree Regression (DTR), Random Forest Regression (RFR), Extra Tree Regression (ETR), and eXtreme Gradient Boosting Regression (XGBoost Regression) ML algorithms are deployed for joint moment prediction. The results reveal EO-ETR as the most effective algorithm for ankle, knee, and neck joint moment prediction, while MPA-ETR exhibits superior performance for hip joint prediction. This approach demonstrates potential for enhancing accuracy in joint moment estimation with minimal feature input, offering implications for biomechanical research and clinical applications.

Efficient prediction of optical properties in hexagonal PCF using machine learning models

This research explores the use of machine learning (ML) models to forecast optical characteristics in photonic crystal fibers (PCF). Specifically, we focus on a solid core index-guided PCF with a hexagonal cladding arrangement. The primary challenges to PCF propagation analysis and predictions are accuracy, computational error, and time constraints. To address these difficulties, we have specially used ML ensemble models including Decision Tree Regressor (DTR), Random Forest Regressor (RFR), Gradient Boosting Regressor (GBR), eXtreme Gradient Boosting Regression (XGBR), and Bagging Regressor (BR). Model performance is assessed using metrics like Mean Squared Error (MSE) and R-squared (R2) through 10-fold cross-validation. Our key findings show that the GBR model outperforms other models and shows extremely low MSE and outstanding R2 values in predicting effective refractive index (Neff), effective mode area (Aeff), confinement loss, and dispersion. In addition, the study compares the performance of ML models with that of previous works using Artificial Neural Network (ANN), demonstrating improved efficiency in predicting optical characteristics for hexagonal PCFs.

Evaluating and Comparing AI Models for Hourly Energy Demand Prediction

This study investigates the application of various AI models to predict energy demand, comparing the performance of four specific models: Decision Tree, Random Forest, Gradient Boosting, and Linear Regression. The evaluation of these models' prediction performance reveals that ensemble methods like Random Forest and Gradient Boosting exhibit promising generalization capabilities, while the Decision Tree model shows high training accuracy but suffers from overfitting. The discussion underscores the importance of ensemble techniques and feature engineering optimization in mitigating overfitting and enhancing forecast accuracy. Furthermore, the study highlights the potential of AI-driven approaches to promote sustainability and resilience in energy systems, emphasizing the need for further optimization and collaboration among stakeholders to achieve a cleaner, more sustainable energy future.

Temporal heterogeneity in the performance of machine learning models for PM2.5 concentration estimation

Machine learning (ML) methods have been applied extensively to simulate air pollutant concentrations and assess individual exposure in epidemiological studies. However, there is still a paucity of research on the temporal heterogeneity of ML model performance and the impact of dataset size. To explore the temporal heterogeneity in model performance when estimating daily concentrations of fine particulate matter (PM2.5) across China in 2021, we compared five decision tree-based ML models (Random Forest (RF), Categorical Boosting (CatBoost), Gradient Boost Regression Tree (GBRT), eXtreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM)) across daily scales within three distinct timeframes. The performance of all models was evaluated using cross-validation. We observed that the performance of ML models varied with time, which showed a significant correlation with PM2.5 concentration. Among the 365 days in 2021, RF model performed best, the annual mean R2 was 0.86, with a minimum of 0.84 and a maximum of up to 0.95. For RF, we chose a cubic polynomial curve to fit the relationship between model performance and PM2.5 concentrations, and based on this, we devised a model selection strategy for different time scales, achieving an accuracy rate of up to 79.45 %, with the selected models having an average R2 of 0.85, and a maximum of up to 0.95. Additionally, we found that increasing the dataset size did not significantly improve model performance. Instead, it resulted in considerably longer runtime and increased memory usage. The methodology and findings of this study hold significant value for advancing the development of more efficient and precise modeling approaches for air pollutant concentrations. Furthermore, this research provides a foundation for regional air pollutant governance and future health-related research.

Prediction of Robotic Anastomosis Competency Evaluation (RACE) metrics during vesico-urethral anastomosis using electroencephalography, eye-tracking, and machine learning

Residents learn the vesico-urethral anastomosis (VUA), a key step in robot-assisted radical prostatectomy (RARP), early in their training. VUA assessment and training significantly impact patient outcomes and have high educational value. This study aimed to develop objective prediction models for the Robotic Anastomosis Competency Evaluation (RACE) metrics using electroencephalogram (EEG) and eye-tracking data. Data were recorded from 23 participants performing robot-assisted VUA (henceforth ‘anastomosis’) on plastic models and animal tissue using the da Vinci surgical robot. EEG and eye-tracking features were extracted, and participants’ anastomosis subtask performance was assessed by three raters using the RACE tool and operative videos. Random forest regression (RFR) and gradient boosting regression (GBR) models were developed to predict RACE scores using extracted features, while linear mixed models (LMM) identified associations between features and RACE scores. Overall performance scores significantly differed among inexperienced, competent, and experienced skill levels (P value < 0.0001). For plastic anastomoses, R2 values for predicting unseen test scores were: needle positioning (0.79), needle entry (0.74), needle driving and tissue trauma (0.80), suture placement (0.75), and tissue approximation (0.70). For tissue anastomoses, the values were 0.62, 0.76, 0.65, 0.68, and 0.62, respectively. The models could enhance RARP anastomosis training by offering objective performance feedback to trainees.

Assessing the impact of climate variability on maize yields in the different regions of Ghana-A machine learning perspective.

Climate variability has become one of the most pressing issues of our time, affecting various aspects of the environment, including the agriculture sector. This study examines the impact of climate variability on Ghana's maize yield for all agro-ecological zones and administrative regions in Ghana using annual data from 1992 to 2019. The study also employs the stacking ensemble learning model (SELM) in predicting the maize yield in the different regions taking random forest (RF), support vector machine (SVM), gradient boosting (GB), decision tree (DT), and linear regression (LR) as base models. The findings of the study reveal that maize production in the regions of Ghana is inconsistent, with some regions having high variability. All the climate variables considered have positive impact on maize yield, with a lesser variability of temperature in the Guinea savanna zones and a higher temperature variability in the Volta Region. Carbon dioxide (CO2) also plays a significant role in predicting maize yield across all regions of Ghana. Among the machine learning models utilized, the stacking ensemble model consistently performed better in many regions such as in the Western, Upper East, Upper West, and Greater Accra regions. These findings are important in understanding the impact of climate variability on the yield of maize in Ghana, highlighting regional disparities in maize yield in the country, and highlighting the need for advanced techniques for forecasting, which are important for further investigation and interventions for agricultural planning and decision-making on food security in Ghana.

Predicting Corporate Bond Illiquidity via Machine Learning

This paper tests the predictive performance of machine learning methods in estimating the illiquidity of US corporate bonds. Machine learning techniques outperform the historical illiquidity-based approach, the most commonly applied benchmark in practice, from both a statistical and an economic perspective. Gradient-boosted regression trees perform particularly well. Historical illiquidity is the most important single predictor variable, but several fundamental and return- as well as risk-based covariates also possess predictive power. Capturing nonlinear effects and interactions among these predictors further enhances forecasting performance. For practitioners, the choice of the appropriate machine learning model depends on the specific application.

Machine Learning-Based Prediction of Hemoglobinopathies Using Complete Blood Count Data.

Hemoglobinopathies, the most common inherited blood disorder, are frequently underdiagnosed. Early identification of carriers is important for genetic counseling of couples at risk. The aim of this study was to develop and validate a novel machine learning model on a multicenter data set, covering a wide spectrum of hemoglobinopathies based on routine complete blood count (CBC) testing. Hemoglobinopathy test results from 10 322 adults were extracted retrospectively from 8 Dutch laboratories. eXtreme Gradient Boosting (XGB) and logistic regression models were developed to differentiate negative from positive hemoglobinopathy cases, using 7 routine CBC parameters. External validation was conducted on a data set from an independent Dutch laboratory, with an additional external validation on a Spanish data set (n = 2629) specifically for differentiating thalassemia from iron deficiency anemia (IDA). The XGB and logistic regression models achieved an area under the receiver operating characteristic (AUROC) of 0.88 and 0.84, respectively, in distinguishing negative from positive hemoglobinopathy cases in the independent external validation set. Subclass analysis showed that the XGB model reached an AUROC of 0.97 for β-thalassemia, 0.98 for α0-thalassemia, 0.95 for homozygous α+-thalassemia, 0.78 for heterozygous α+-thalassemia, and 0.94 for the structural hemoglobin variants Hemoglobin C, Hemoglobin D, Hemoglobin E. Both models attained AUROCs of 0.95 in differentiating IDA from thalassemia. Both the XGB and logistic regression model demonstrate high accuracy in predicting a broad range of hemoglobinopathies and are effective in differentiating hemoglobinopathies from IDA. Integration of these models into the laboratory information system facilitates automated hemoglobinopathy detection using routine CBC parameters.

Machine learning approach to predict Hansen solubility parameters of cocrystal coformers via integrating group contribution and COSMO-RS

Hansen Solubility Parameters (HSPs) have been a hot topic on predicting the tendency of pharmaceutical cocrystals formation and cocrystal coformers (CCFs) screening. However, the limitation of such application is the lack of models to accurately predict the values of HSPs for drug CCFs with more structural complexity. Accordingly, three ML (machine learning) models, i.e. ANN (Artificial Neural Network), XGBoostRegressor (Extreme Gradient Boosting Regressor) and LGBMRegressor (Light Gradient Boosting Machine Regressor), were developed for predicting the HSPs on CCFs screening for drugs. The HSPs database for 181 CCFs (containing alcohols, alkenes, aromatics, haloalkanes, amines, ketones, ethers, amides, esters, pharmaceuticals, alkanes, acids, nitroalkanes) were established and classified into the training set (140 compounds) and the test set (41 compounds with various functional polarity and groups, covering solid reagents and solvents). The prediction molecular descriptors were combined from the GC (Group Contribution) methods, the COSMO-RS (the Conductor-like Screening Model for Real Solvents) sigma-moments and energy descriptors. The results showed that ANN and XGBoostRegressor beat out LGBMRegressor in predicting HSPs for CCFs. Finally, SHapley Additive exPlanations (SHAP) was employed to visualize and explain the most important characteristics and effects on predicting HSPs via XGBoostRegressor, indicating that CH3, M2 and MHbdon3 had a significant influence and high contribution to the prediction of δd, δp and δh, respectively. The coupled GC and COSMO-RS strategy had been proven as a promising tool to predict HSPs through XGBoostRegressor for screening, designing, and selecting CCFs for drugs.

Machine learning prediction of bio-polyol yields and hydroxyl values from acid-catalyzed liquefaction of lignocellulosic biomass

Amid the escalating demand for alternatives to petroleum resources and the imperative to decrease carbon emissions, there is an increasing interest in transforming lignocellulosic biomass into valuable chemicals. This study utilized machine learning models to analyze the acid-catalyzed liquefaction process of lignocellulosic biomass and to predict and characterize the yield and hydroxyl value (HV) of bio-polyols. A total of 612 yield data samples and 229 HV data samples were collected and analyzed. From these data, four machine learning models were constructed: Random Forest (RF), Gradient Boosting Regression (GBR), Gaussian Process Regression (GPR), and Support Vector Regression (SVR). The Bayesian Optimization Algorithm (BOA) was applied to refine the hyperparameters of the models, thereby enhancing their predictive accuracy. To clarify the effects of input features on predictive outcomes and to understand their interactive dynamics, one-dimensional and two-dimensional Accumulated Local Effects (ALE) analyses were performed. The results show that the GBR model was particularly precise in forecasting the yield and HV of bio-polyols, with corresponding test set determination coefficients (R2) of 0.82 and 0.91. The most influential factors on yield were reaction time (15.1 %), lignin content (13.1 %), and glycerol mass (12 %), while for HV, they were glycerol mass (28.9 %), particle size (15.1 %), and liquid-to-solid ratio (11.1 %). The ALE analysis also exposed the intricate interaction mechanisms between feedstock composition and liquefaction conditions. The verification results confirmed that the optimal model displayed exceptional generalization capabilities, with a Mean Absolute Percentage Error (MAPE) of 9.18 %. Utilizing this model, a user-friendly software package has been developed to enable rapid and precise prediction of the lignocellulosic biomass conversion process. This research not only delivers strategies to cut down on experimental costs and time but also offers a novel perspective on the industrial utilization of biomass-derived polyols.

Predicting trajectories of temperate forest understorey vegetation responses to global change

Predicting forest understorey community responses to global change and forest management is vital given the importance of the understorey for biodiversity conservation and forest functioning. Though substantial effort has gone into disentangling the impact of global change on understorey communities, scarcity of information on site-specific environmental drivers across large temporal-spatial scales has limited our ability to predict global change effects at specific forest sites. In this study, using vegetation resurvey and soil data from 1363 plots across temperate Europe, we applied a machine learning approach (gradient boosting regression, GBR) to model and predict site-specific responses of four understorey properties to global change. We applied our final GBR models at 8 forest sites in Austria to validate the model performance, predict understorey trajectories, and evaluate the effect of alternative scenarios for future nitrogen(N) deposition, climate change and forest management on the projected trajectories. Our results showed that the R² value of the four final GBR models on the independent testing dataset ranged between 0.611 and 0.723 and the most important environmental drivers in predicting the trajectory of understorey properties at specific forest sites were soil pH, soil total carbon-to-nitrogen ratio, overstorey shade-casting ability and regional-scale mean annual precipitation. The out-of-sample R2 value of the four final GBR models on the Austrian data ranged between 0.224 and 0.561. The forecasted trajectories for the Austrian forest sites showed that site-specific understorey responses to near-future climate warming were expected to be weak. Under N deposition decreases, the proportion of woody species was predicted to increase, while species richness and total vegetation cover were predicted to decrease. Furthermore, under a closed canopy, the understorey community was predicted to shift towards more woody species and more forest specialists, albeit with reduced species richness and vegetation cover. Given expected warming and declining N pollution pressures, our presented GBR models allow the prediction of trajectories of understorey vegetation responses to global change and management interventions at specific forest sites. Such projections could aid forest management in addressing challenges posed by global change.

An artificial intelligence framework for predicting operational energy consumption in office buildings

This research delves into the intricate dynamics of energy consumption in buildings, using data-driven modeling with machine learning (ML) algorithms to optimize design choices for heightened energy efficiency. A substantial dataset, comprising 66,800 operational energy records from office buildings in 40 cities worldwide representing 10 distinct climate conditions, was analyzed using four ML algorithms: Random Forest Regressor, Extra Trees Regressor, Gradient Boosting Regressor (GBR), and eXtreme Gradient Boosting Regressor (XGBR). Notably, the developed GBR and XGBR models demonstrated superior precision in predicting energy use intensity during the testing phase, achieving a mean absolute percentage error of less than 1.2 % and a coefficient of determination of more than 0.99. The findings highlighted the significant impact of building shape and temperature on annual operational energy consumption, providing valuable insights for energy-efficient design optimization. Finally, the grey wolf optimizer was applied to identify optimal design parameters for various scenarios, laying the groundwork for designing energy-efficient buildings in cities with diverse meteorological patterns.

Estimation of invasive coronary perfusion pressure using electrocardiogram and Photoplethysmography in a porcine model of cardiac arrest

BackgroundCoronary perfusion pressure (CPP) indicates spontaneous return of circulation and is recommended for high-quality cardiopulmonary resuscitation (CPR). This study aimed to investigate a method for non-invasive estimation of CPP using electrocardiography (ECG) and photoplethysmography (PPG) during CPR. MethodsNine pigs were used in this study. ECG, PPG, invasive arterial blood pressure (ABP), and right atrial pressure (RAP) signals were simultaneously recorded. The CPPs were estimated using three datasets: (a) ECG, (b) PPG, and (c) ECG and PPG, and were compared with invasively measured CPPs. Four machine-learning algorithms, namely support vector regression, random forest (RF), K-nearest neighbor, and gradient-boosted regression tree, were used for estimation of CPP. ResultsThe RF model with a combined ECG and PPG dataset achieved better estimation of CPP than the other algorithms. Specifically, the mean absolute error was 4.49 mmHg, the root mean square error was 6.15 mm Hg, and the adjusted R2 was 0.75. A strong correlation was found between the non-invasive estimation and invasive measurement of CPP (r = 0.88), which supported our hypothesis that machine-learning-based analysis of ECG and PPG parameters can provide a non-invasive estimation of CPP for CPR. ConclusionsThis study proposes a novel estimation of CPP using ECG and PPG with machine-learning-based algorithms. Non-invasively estimated CPP showed a high correlation with invasively measured CPP and may serve as an easy-to-use physiological indicator for high-quality CPR treatment.

Development of machine learning based model for low-temperature PEM fuel cells

Low-Temperature Proton Exchange Membrane Fuel Cells (LT-PEMFC) are favored as an alternative power source due to their high efficiency, rapid initialization, shut-down cycles, and zero emissions. Developing an effective model for LT-PEMFC is essential. In this study, machine learning models are created for LT-PEMFC, utilizing techniques such as Gradient Boosting Regression (GBR), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM) to predict cell voltage based on operating parameters. The dataset is generated using an in-house physics-based MATLAB model, complemented by experimental data from elsewhere. GBR exhibits superiority over XGBoost, LightGBM, and RF. These data-based models for LT-PEMFC, developed on generated datasets, achieve R2≥ 0.99 and MAPE ≤ 0.06 during testing. These models are further validated on experimental data with R2≥ 0.90 and MAPE ≤ 0.1. This underscores the ability to construct accurate data-based models and thus reducing reliance on extensive experimentation.