Gradient Boosting Regression Research Articles

A machine learning model for predicting recovery rates of COVID-19 patients

During disease epidemics, any trial to improve healthcare systems entails preserving lives. Therefore, predicting which patients are at high risk becomes critical and challenging when confronted with a novel virus. The recent COVID-19 changed many people’s perspectives on how to approach diseases. According to the lack of medical resources, it is important to identify the patients who need instant medical care. This research proposes a machine learning model to identify high-risk patients that require specific medical attention. Specifically, extreme gradient boosting (XGboost), random forest (RF), and logistic regression (LR) are used in the ensemble method to classify COVID-19 patients at high risk. The dataset consists of 361 medical records for severe COVID-19 patients which have included 195 survivors. The most correlated features (neutrophils (%), hypersensitive c-reactive protein, lactate dehydrogenase, age, procalcitonin, and neutrophils count) are selected to be used in classification. Different machine learning classifiers are applied to the mentioned dataset to find out the optimum classifiers to be used in the ensemble method. 98% is the most optimal accuracy achieved with the proposed model.

Forecasting next-hour electricity demand in small-scale territories: Evidence from Jordan

In exceptional times of wars, natural crises (e.g., snow storms), or hosting massive events (e.g., international sports events), prior knowledge of hour-by-hour electricity demand might become critical for the concerned areas. Anticipating the next-hour demand within a bounded location might help cope with challenging situations. In this paper, we are concerned with the problem of electricity demand forecasting for the next hour within a small area in the context of the country of Jordan relying exclusively on historical electricity consumption. Existing electricity demand forecasting models predict electricity demand for the next day for the whole country of Jordan or for large cities using numerous indicators such as weather conditions and season-related features, together with historical electricity consumption. Our proposed solution consists of preprocessing the hourly-electricity-consumption data to rearrange it into time series of 25-hour length. Then, we applied supervised machine learning algorithms to build predictive models for electricity demand. Our methods proved effective to a certain degree in forecasting electricity demand using time series of historical electricity consumptions only. We employ several predictive machine learning models: neural network, decision tree, random forest, AdaBoost, support vector regression, nearest neighbors, linear regression, bayesian ridge, partial least squares, gradient boosting, and principal component regression. We found that the most critical indicators in our predictive models are the electricity consumption of the two hours preceding the target hour and the electricity consumption the day before in the hour (the same as the target hour) and the next hour. Our predictive models can efficiently forecast electricity demand in a small rural area in Jordan with an accuracy of nearly 97% at best and 92% at least. Our predictive models are constructed using historical data that includes the electricity consumption of a small Jordanian district for the years 2019–2021.

Application of machine learning approaches for modelling crack growth rates

AbstractThe scientific community has widely accepted the use of machine learning techniques to tackle complex engineering problems. Among the most intriguing problems is finding the correlation between alloy steel properties and cyclic fatigue and crack growth rate. Employing machine‐learning models can provide more robust and accurate predictive models to address such challenges. This paper presents the application of four machine learning models, namely decision trees (DT), random forest (RF), adoptive boosting (AdaBoost), and gradient boosting regression tree (GBRT) to predict the crack growth rate of steel/alloys. The study utilizes a large database gathered from literature to construct the predictive models and compares the results using various statistical metrics and graphical representation. The study's findings demonstrate the effectiveness and suitability of machine learning techniques to handle complex databases related to fatigue problems.

Price Prediction Using Web Scraping and Machine Learning Algorithms in the Used Car Market

The development of technology increases data traffic and data size day by day. Therefore, it has become very important to collect and interpret data. This study, it is aimed to analyze the car sales data collected using web scraping techniques by using machine learning algorithms and to create a price estimation model. The data needed for analysis was collected using Selenium and BeautifulSoup and prepared for analysis by applying various data preprocessing steps. Lasso regression and PCA analysis were used for feature selection and size reduction, and the GridSearchCV method was used for hyperparameter tuning. The results were evaluated with machine learning algorithms. 
 Random Forest, K-Nearest Neighbor, Gradient Boost, AdaBoost, Support Vector and XGBoost regression algorithms were used in the analysis. The obtained analysis results were evaluated together with Mean Square Error (MSE), Root Mean Square Error (RMSE) and Coefficient of Determination (R-square). When the results for data set 1 were examined, the model that gave the best results was XGBoost Regression with 0.973 R2, 0.026 MSE and 0.161 RMSE values. When the results for data set 2 were examined, the model that gave the best results was K-Nearest Neighbor Regression with 0.978 R2, 0.021 MSE and 0.145 RMSE values.

Machine learning-based prediction of optical band gap in WO3 and its derivatives for semiconducting applications

Band gap engineering plays a vital role in designing a functional semiconductor for specific applications in the arena of material science. Recently, machine learning algorithms are turning out very useful tools for the prediction of the optical band gaps of transition metal oxides (TMO). TMO are the most commonly used semiconductors for a wide range of applications, and Tungsten Oxide (WO3) is one of the potential candidates of the TMO family. Herein, we have reported an alternative way to determine the bandgap of WO3 and its derivatives for the fabrication of novel semiconducting devices. Conventional methods; mainly based on density functional theory (DFT) and UV-Vis Spectroscopy sometimes suffers from higher computational and experimentation costs respectively. To avoid such many other complications, we have tested ten machine learning algorithms and further verified that gradient boost regression (GBR) turned out to be the most suitable algorithm which exhibited root-mean-square error of around 7.99E-05 eV and an R2 score near 1 (0.9999). The exceptional precision of the regression model is achieved by relying solely on structural descriptors and training the model with a dataset that exclusively comprises experimentally measured band gaps. Estimation of the bandgaps with the help of such machine learning algorithms is practicable for large set virtual screening of WO3 and WO3-based derivatives for the range of applications in the field of energy storage, optoelectronics, sensing, and many other semiconducting applications.

Accident Risk Prediction through Machine Learning: A Comprehensive Study

Abstract—In the realm of road safety, the mitigation of accidents and their potential consequences remains a paramount concern. This research paper embarks on an extensive exploration of accident risk prediction, harnessing the capabilities of Extra Trees Regressor and XGBoost for the critical task of parameter importance analysis. The study is centered around the development of a sophisticated web-based prediction system, encompassing various facets such as system architecture, meticulous data preprocessing, and the intricate intricacies of machine learning model development. The dataset used for the same is from Kaggle comprising of data about accidents in US between 2016-2023 . Through the lens of parameter importance analysis, this research unveils the key determinants that underpin accident risk, casting a spotlight on the profound implications of these insights on prediction accuracy which is currently 85%. The ensuing discourse delves into the nuanced interpretation of results, culminating with a forward- looking perspective that outlines potential pathways for future research endeavors in this consequential domain. Keywords—Accident Risk Prediction, Accident Prevention, Road Safety, Data-Driven Analysis, Feature Selection, Risk Assessment, Machine Learning Algorithms, Data Preprocessing, Model Evaluation, Hist Gradient Boosting Regressor, XGBoost, Parameter Importance Analysis, Predictive Systems.

Continuous blood pressure monitoring using photoplethysmography and electrocardiogram signals by random forest feature selection and GWO-GBRT prediction model

Blood pressure (BP) is an important indicator of a person's cardiovascular status, but accurate, continuous BP monitoring remains a challenge. To improve the monitoring performance of continuous BP using electrocardiogram (ECG) and photoplethysmography (PPG) signals, a continuous BP monitoring method based on random forest feature selection (RFFS) and a gray wolf optimization-gradient boosting regression tree (GWO-GBRT) prediction model is developed in this study. First, features in time, frequency, and time–frequency domains are extracted from PPG and ECG signals to supply comprehensive characteristics for BP monitoring. Then, the RFFS method is employed to select sensitive features highly correlated with BP from the candidate feature sets, which reduced redundant information and improved monitoring efficiency. Next, a hybrid prediction model of the gray wolf optimization (GWO) technique and gradient boosting regression tree (GBRT) algorithm is established to learn the dependency relationship between BP and sensitive features. And a new fitness function of GWO is designed to balance the monitoring accuracy and consistency. Finally, ablation and comparative experiments demonstrate the effectiveness and advancement of the proposed method, using the ECG and PPG signals of 150 people downloaded from the MIMIC-III database. The mean absolute error (MAE) and standard deviation (STD) of the proposed method in predicting systolic blood pressure were 2.91 mmHg and 3.97 mmHg, and those of diastolic blood pressure were 1.71 mmHg and 2.35 mmHg. Its monitoring accuracy has surpassed the Association for Advancement of Medical Instrumentation (AAMI) standard and reached the “A” level standard of the British Hypertension Society (BHS) protocol.

Groundwater Level Trend Analysis and Prediction in the Upper Crocodile Sub-Basin, South Africa

Disasters related to climate change regarding water resources are on the rise in terms of scale and severity. Therefore, predicting groundwater levels (GWLs) is a crucial means to aid adaptive capacity towards disasters related to climate change in water resources. In this study, a Gradient Boosting (GB) regression modelling approach for GWL prediction as a function of rainfall and antecedent GWL is used. A correlation analysis carried out from 2011 to 2020 demonstrated that monthly GWLs can be predicted by antecedent GWLs and rainfall. The study also sought to understand the long-term effects of climate events on groundwater levels over the study area through a Mann–Kendall (MK) trend analysis. A total of 50% of the groundwater stations revealed declining trends, while 25% had no trends and the other 25% showed an increasing trend. Again, the correlation analysis results were used in justifying the trends. The GB predictive model performed satisfactorily for all groundwater stations, with the MSE values ranging from 0.03 to 0.304 and the MAE varying from 0.12 to 0.496 in the validation period. The R2 ranged from 0.795 to 0.902 for the overall period. Therefore, based on projected rainfall and antecedent groundwater levels, future GWLs can be predicted using the GB model derived in this study.

Machine learning-based predictions of yield strength for neutron-irradiated ferritic/martensitic steels

Due to their superior resistance to radiation-induced damage, Ferritic/Martensitic (F/M) steels are promising structural materials for advanced nuclear reactors. This study investigates machine learning (ML) methodologies to predict the yield strength of neutron-irradiated F/M steels. Popular ML algorithms, such as Random Forest (RF), Extreme Gradient Boosting (XGBOOST), Gradient Boosting (GBOOST), and Support Vector Regression (SVR), were trained on a experimental dataset to understand the relationship between the input variables (e.g., irradiation dose, irradiation temperature, tensile test condition, heat treatment conditions and steels composition) and the output variable (yield strength). Further, the SHapley Additive exPlanations (SHAP) algorithm was employed to obtain the importance hierarchy of the input variables for their selection. Post-training and testing ML algorithms, their performance was evaluated by assessing their ability to predict the unseen/validation dataset. Among the algorithms tested, XGBOOST demonstrated the highest performance in predicting the validation dataset, followed by RF, GBOOST, and SVR. Synthetic experiments show that the trained ML algorithms can capture the trends between the irradiation input variables and the yield strength. Overall, the trained ML algorithms overcame challenges such as data uncertainties, smaller and sparser datasets, and other complexities and predicted almost 70% to 75% of the datapoints in the unseen/validation dataset within one standard deviation.

Modeling of the control logic of a UASS based on coefficient of variation spraying distribution analysis in an indoor flight simulator.

In the past decade, unmanned aerial spraying systems (UASS) have emerged as an effective crop treatment platform option, competing with other ground vehicle treatments. The development of this platform has provided an effective spraying system that can be used on all crop types and in all weather conditions. However, related research has not been able to develop a UASS that can be operated in windy conditions with a low drift percentage. In this research, spraying was simulated in an indoor flight simulator by considering flight speed, altitude, wind speed, wind direction, rotor rotation, interval, spraying pattern, and nozzle type, which were used as the parameters affecting the output value of the coefficient of variation (CV) of spraying. These parameters were referenced as properties that occur in the field, and using machine learning methods, the CV value was used as a dataset to develop a model that can execute pump opening by controlling the flow rate. There are four machine learning methods used, i.e. random forest regression, gradient boosting, ada boost, and automatic relevance determination regression which are compared with simple linear regression and ridge regression as linear regression. The results revealed that the random forest regression model was the most accurate, with R2 of 0.96 and root mean square error (RMSE) of 0.04%. The developed model was used to simulate spraying with pump opening A, which connects two nozzles in front, and pump opening AB, which connects all four nozzles. Using the logic based on CV value and pesticide quantity, the model can execute the pump opening against the environment and UASS operation.

Accelerating the Discovery of Hybrid Perovskites with Targeted Band Gaps via Interpretable Machine Learning.

The band gap of hybrid organic-inorganic perovskites (HOIPs) is a key factor affecting the light absorption characteristics and thus the performance of perovskite solar cells (PSCs). However, band gap engineering, using experimental trial and error and high-throughput density functional theory calculations, is blind and costly. Therefore, it is critical to statistically identify the multiple factors influencing band gaps and to rationally design perovskites with targeted band gaps. This study combined feature engineering, the gradient-boosted regression tree (GBRT) algorithm, and the genetic algorithm-based symbolic regression (GASR) algorithm to develop an interpretable machine learning (ML) strategy for predicting the band gap of HOIPs accurately and quantitatively interpreting the factors affecting the band gap. Seven best physical features were selected to construct a GBRT model with a root-mean-square error of less than 0.060 eV, and the most important feature is the electronegativity difference between the B-site and the X-site (χB-X). Further, a mathematical formula (Eg = χB-X2 + 0.881χB-X) was constructed with GASR for a quantitative interpretation of the band gap influence patterns. According to the ML model, the HOIP MA0.23FA0.02Cs0.75Pb0.59Sn0.41Br0.24I2.76 was obtained, with a suitable band gap of 1.39 eV. Our proposed interpretable ML strategy provides an effective approach for developing HOIP structures with targeted band gaps, which can also be applied to other material fields.

Rainfall-Runoff modelling using SWAT and eight artificial intelligence models in the Murredu Watershed, India.

The growing concerns surrounding water supply, driven by factors such as population growth and industrialization, have highlighted the need for accurate estimation of streamflow at the river basin level. To achieve this, rainfall-runoff models are widely employed as valuable tools in watershed management. For this specific study, two modelling approaches were employed: the Soil and Water Assessment Tool (SWAT) model and a set of eight artificial intelligence (AI) models. The AI models consisted of seven data-driven approaches, namely k-nearest neighbour regression, support vector regression, linear regression, artificial neural networks, random forest regression, XGBoost, and Histogram-based Gradient Boost regression. Additionally, a deep learning model known as Long Short-Term Memory (LSTM) was also utilized. The study focused on monthly streamflow modelling in the Murredu River basin, with a calibration period from 1999 to 2003 and a validation period from 2004 to 2005, spanning a total of 7years from 1999 to 2005. The results indicated that all nine models were generally suitable for simulating the rainfall-runoff process, with the LSTM model demonstrating exceptional performance in both the calibration (R2 is 0.97 and NSE is 0.96) and validation (R2 is 0.97 and NSE is 0.92) periods. Its high coefficient of determination (R2) and Nash-Sutcliffe efficiency (NSE) values indicated its superior ability to accurately model the rainfall-runoff relationship. While the other models also produced satisfactory results, the findings suggest that selecting the most efficient model, such as the LSTM model, could significantly contribute to the effective management and planning of sustainable water resources in the Murredu watershed.

Machine learning-enabled analysis of product distribution and composition in biomass-coal co-pyrolysis

Co-pyrolysis of biomass and coal presents a promising opportunity for large-scale biomass utilization while reducing fossil fuel consumption. However, this process is highly complex and influenced by various factors, including the physicochemical properties of biomass and coal, their blending ratio, and operating parameters. To effectively comprehend and optimize the biomass-coal co-pyrolysis process, many experiments are required to achieve the desired product quantity and quality. In addressing this challenge, machine learning technology emerges as a valuable solution, as it can learn the relationships between input and output variables from available examples without explicit knowledge of the underlying mechanisms. This study conducts a comprehensive literature review to establish an extensive database encompassing biomass-coal compositions and pyrolysis reaction conditions. Using the collected data, robust statistical analyses are applied to understand better the underlying mechanisms governing the biomass-coal co-pyrolysis process. Four machine learning methods, namely support vector regression, artificial neural network, random forest regression, and gradient boosting regression, are employed to model the process. The best-performing model is subjected to an extensive feature importance analysis to identify the essential input features associated with each output response. The gradient boosting regression technique demonstrates superior performance with excellent results, characterized by higher coefficients of determination (R2 > 0.96) and lower errors (RMSE < 3.01 and MAE < 2.27). The temperature range of 450 to 550 °C, biomass blending ratios between 30 and 60%, and heating rates of 30 to 50 °C/min were identified as the conditions that maximize pyrolysis oil yield. Furthermore, the feature importance analysis reveals that the operating temperature and the biomass blending ratio are the most significant descriptors in the biomass-coal co-pyrolysis process.

Rapid Estimation of Contact Stresses in Imageless Total Knee Arthroplasty

Total knee arthroplasty (TKA) is a surgical technique to replace damaged knee joints with artificial implants. Recently, the imageless TKA has brought a revolutionary improvement to the accuracy of implant placement and ease of surgical process. Based on key anatomical points on the knee, the software guides the surgeon during the TKA procedure. However, the number of revision surgeries is increasing due to malalignment caused by registration error, resulting in imbalanced contact stresses that lead to failure of the TKA. Conventional stress analysis methods involve time-consuming and computationally demanding finite element analysis (FEA). In this work, a machine-learning-based approach estimates the contact pressure on the TKA implants. The machine learning regression model has been trained using FEA data. The optimal preprocessing technique was confirmed by the data without preprocessing, data divided by model size, and data divided by model size and optimal angle. Extreme gradient boosting, random forest, and extra trees regression models were trained to determine the optimal approach. The proposed method estimates the contact stress instantly within 10 percent of the maximum error. This has resulted in a significant reduction in computational costs. The efficiency and reliability of the proposed work have been validated against the published literature.

Modeling and forecasting electricity consumption amid the COVID-19 pandemic: Machine learning vs. nonlinear econometric time series models

Accurately modeling and forecasting electricity consumption remains a challenging task due to the large number of the statistical properties that characterize this time series such as seasonality, trend, sudden changes, slow decay of autocorrelation function, among many others. This study contributes to this literature by using and comparing four advanced time series econometrics models, and four machine learning and deep learning models11These models include the autoregressive model with seasonality, autoregressive models with exogenous variables, the autoregressive fractionally integrated moving average model with exogenous variables, the three state autoregressive Markov switching model with exogenous variable, Prophet, EXtreme Gradient Boosting, Long-Short-Term Memory and Support Vector Regression. to analyze and forecast electricity consumption during COVID-19 pre-lockdown, lockdown, releasing-lockdown, and post-lockdown phases. Monthly data on Qatar’s total electricity consumption has been used from January 2010 to December 2021. The empirical findings demonstrate that both econometric and machine learning models are able to capture most of the important statistical features characterizing electricity consumption. In particular, it is found that climate change based factors, e.g temperature, rainfall, mean sea-level pressure and wind speed, are key determinants of electricity consumption. In terms of forecasting, the results indicate that the autoregressive fractionally integrated moving average and the three state autoregressive Markov switching models with exogenous variables outperform all other models. Policy implications and energy-environmental recommendations are proposed and discussed.

A comparison of the performance of SWAT and artificial intelligence models for monthly rainfall–runoff analysis in the Peddavagu River Basin, India

Abstract Rainfall–runoff (R–R) analysis is essential for sustainable water resource management. In the present study focusing on the Peddavagu River Basin, various modelling approaches were explored, including the widely used Soil and Water Assessment Tool (SWAT) model, as well as seven artificial intelligence (AI) models. The AI models consisted of seven data-driven models, namely support vector regression, artificial neural network, multiple linear regression, Extreme Gradient Boosting (XGBoost) regression, k-nearest neighbour regression, and random forest regression, along with one deep learning model called long short-term memory (LSTM). To evaluate the performance of these models, a calibration period from 1990 to 2005 and a validation period from 2006 to 2010 were considered. The evaluation metrics used were R2 (coefficient of determination) and NSE (Nash–Sutcliffe Efficiency). The study's findings revealed that all eight models yielded generally acceptable results for modelling the R–R process in the Peddavagu River Basin. Specifically, the LSTM demonstrated very good performance in simulating R–R during both the calibration period (R2 is 0.88 and NSE is 0.88) and the validation period (R2 is 0.88 and NSE is 0.85). In conclusion, the study highlighted the growing trend of adopting AI techniques, particularly the LSTM model, for R–R analysis.

Machine learning and Shapley Additive Explanation-based interpretable prediction of the electrocatalytic performance of N-doped carbon materials

Enhancing the kinetic rate of cathodic oxygen reduction reaction (ORR) by catalysts is the key to improve the performance of microbial fuel cells (MFCs). Metal-free ORR catalysts represented by nitrogen-doped carbon materials have been extensively investigated and have shown excellent catalytic effects for oxygen reduction reaction. However, it is difficult to clarify the coupling effect physicochemical properties of nitrogen-doped carbon materials on their catalytic effect (i.e. electricity production performance of MFCs) by traditional experimental methods. Therefore, in this study, six machine learning models were combined with SHAP to develop prediction models for the power density ratio of MFCs for reflecting the catalytic performance of nitrogen-doped carbon materials by using physicochemical properties, such as elemental composition, functional group structure, and pore structure, as input features. The gradient boosting regression (GBR) model was found to have the highest prediction accuracy on the test set, with R2 and RMSE of 0.86 and 0.09, respectively. The SHAP method was used to interpret the output of the GBR model and reveal the mechanism of interaction between different characteristic variables. It was found that pyridine nitrogen is the most important input characteristic of the nitrogen elements, as its corresponding SHAP value reaches above 0.1. Surprisingly, an increase in the content of oxygen significantly attenuates the extent to which changes in nitrogen content affect the system, although its effect on power density prediction is inconsiderable. In addition, the optimal range of important physicochemical properties of nitrogen-doped carbon materials was obtained by the SHAP method. The carbon materials have relatively high catalytic performance under the conditions of C(at%) between 85% and 90%, N(at%) > 2.5%, Pyridine-N(at%) > 30%, Vtotal between 0.3 and 0.6 cm3g−1, and SBET > 1000 m2g−1. This study can provide theoretical guidance for subsequent experimental design of carbon-based nitrogen-doped electrocatalysts.

Estimation of grassland height using optical and SAR remote sensing data

Grassland height is an important indicator used to evaluate ecological environments in grasslands. The study explored the potential of optical and SAR (synthetic aperture radar) remote sensing data in estimating grassland height with machine learning methods and constructed the grassland height remote sensing inversion model. Then the mean grassland height in August 2015 was estimated with the inversion model in Inner Mongolia Autonomous Region, China. Among classification and regression tree (CART), random forest (RF), and gradient boosting regression tree (GBRT) models, the GBRT model had the highest estimation accuracy. There was little correlation between SAR data and grassland height, and SAR data produced poor accuracy for grassland height estimation. The grassland height could be better estimated by optical remote sensing data. The best estimation accuracy (GBRT model: for training data: R2 = 0.71, RMSE = 3.58 cm, P < 0.01; for test data: R2 = 0.58, RMSE = 3.94 cm, P < 0.01) was achieved by the combination of optical and SAR remote sensing data. However, SAR data played an auxiliary role, and the estimation of grassland height was mainly realized by optical data. The average vegetation height of the grasslands in Inner Mongolia in August 2015 was 19.15 cm, gradually decreasing from northeast to southwest. The study proposed a high-precision method for estimating grassland height, which provided a basis for studying grassland environments and conditions.

Improved gray correlation analysis and combined prediction model for aviation accidents

PurposeThe purpose of this paper is to identify the key influencing factors of aviation accidents and to predict the aviation accidents caused by the factors.Design/methodology/approachThis paper proposes an improved gray correlation analysis (IGCA) theory to make the relational analysis of aviation accidents and influencing factors and find out the critical causes of aviation accidents. The optimal varying weight combination model (OVW-CM) is constructed based on gradient boosted regression tree (GBRT), extreme gradient boosting (XGBoost) and support vector regression (SVR) to predict aviation accidents due to critical factors.FindingsThe global aviation accident data from 1919 to 2020 is selected as the experimental data. The airplane, takeoff/landing and unexpected results are the leading causes of the aviation accidents based on IGCA. Then GBRT, XGBoost, SVR, equal-weight combination model (EQ-CM), variance-covariance combination model (VCW-CM) and OVW-CM are used to predict aviation accidents caused by airplane, takeoff/landing and unexpected results, respectively. The experimental results show that OVW-CM has a better prediction effect, and the prediction accuracy and stability are higher than other models.Originality/valueUnlike the traditional gray correlation analysis (GCA), IGCA weights the sample by distance analysis to more objectively reflect the degree of influence of different factors on aviation accidents. OVW-CM is built by minimizing the combined prediction error at sample points and assigns different weights to different individual models at different moments, which can make full use of the advantages of each model and has higher prediction accuracy. And the model parameters of GBRT, XGBoost and SVR are optimized by the particle swarm algorithm. The study can guide the analysis and prediction of aviation accidents and provide a scientific basis for aviation safety management.

Predicting 7-day unplanned readmission in elderly patients with coronary heart disease using machine learning.

Short-term unplanned readmission is always neglected, especially for elderly patients with coronary heart disease (CHD). However, tools to predict unplanned readmission are lacking. This study aimed to establish the most effective predictive model for the unplanned 7-day readmission in elderly CHD patients using machine learning (ML) algorithms. The detailed clinical data of elderly CHD patients were collected retrospectively. Five ML algorithms, including extreme gradient boosting (XGB), random forest, multilayer perceptron, categorical boosting, and logistic regression, were used to establish predictive models. We used the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, the F1 value, the Brier score, the area under the precision-recall curve (AUPRC), and the calibration curve to evaluate the performance of ML models. The SHapley Additive exPlanations (SHAP) value was used to interpret the best model. The final study included 834 elderly CHD patients, whose average age was 73.5 ± 8.4 years, among whom 426 (51.08%) were men and 139 had 7-day unplanned readmissions. The XGB model had the best performance, exhibiting the highest AUC (0.9729), accuracy (0.9173), F1 value (0.9134), and AUPRC (0.9766). The Brier score of the XGB model was 0.08. The calibration curve of the XGB model showed good performance. The SHAP method showed that fracture, hypertension, length of stay, aspirin, and D-dimer were the most important indicators for the risk of 7-day unplanned readmissions. The top 10 variables were used to build a compact XGB, which also showed good predictive performance. In this study, five ML algorithms were used to predict 7-day unplanned readmissions in elderly patients with CHD. The XGB model had the best predictive performance and potential clinical application perspective.