Performance Of Prediction Models Research Articles

Is cardiovascular risk profiling from UK Biobank retinal images using explicit deep learning estimates of traditional risk factors equivalent to actual risk measurements? A prospective cohort study design

ObjectiveDespite extensive exploration of potential biomarkers of cardiovascular diseases (CVDs) derived from retinal images, it remains unclear how retinal images contribute to CVD risk profiling and how the results can...

Application of machine learning and deep learning techniques in modeling the associations between air pollution and meteorological parameters in urban areas of tehran metropolis.

Tehran, the most crowded city in Iran, suffers from severe air pollution, particularly during the cold months. This research endeavored to examine the statistical relationships between criteria air pollutants (CO, NO2, SO2, O3, PM10, and PM2.5) and meteorological elements (temperature, rainfall, wind speed, relative humidity, air pressure, sunshine hours, solar radiation, and cloudiness), as well as assess and compare the efficacy of six different algorithms (multiple linear regression (MLR), generalized additive model (GAM), classification and regression trees (CART), random forest (RF), gradient boosting machine (GBM), and deep learning (DL)) in modeling pollutants and climatic factors responsible for variations in Tehran's air pollution levels from 2001 to 2021 using R 4.3.2 software. The results of this study showed that O3 was strongly affected by weather conditions, while other pollutants were mainly influenced by each other than by meteorological parameters and more extensive research is required to pinpoint the precise impact of human activity on these pollutant levels in Tehran. Also based on the predictive model performance evaluation and concerning the principle of parsimony, in half of the cases, the MLR outperformed other models, despite its seeming simplicity and principal assumptions dependence. In other situations, the GAM was a good substitute.

Neutral Benchmarking of Survival Models in Health Sciences: Comparative Study of Classical and Machine Learning Techniques

Survival analysis plays a central role in diverse research fields, especially in health sciences. As an analytical tool, it can be used to help improve patients’ survival time, or at least, reduce the prospects of recurrence in cancer studies. However, approaches to the predictive performance of the current survival models mainly center on clinical data along with the classical survival methods. For censored “omics” data, the performance of survival models has not been thoroughly studied, either often due to their high dimensionality issues or reliance on binarizing the survival time for classification analysis. We aim to present a neural benchmark approach that analyzes and compares a broad range of classical and state-of-the-art machine learning survival models for “omics” and clinical datasets. All the methods considered in our study are evaluated using predictability as a performance measure. The study is systematically designed to make 36 comparisons (9 methods over 4 datasets, i.e., 2 clinical and 2 omics), and shows that, in practice, predictability of survival models does vary across real-world datasets, model choice, as well as the evaluation metric. From our results, we emphasize that performance criteria can play a key role in a balanced assessment of diverse survival models. Moreover, the Multitask Logistic Regression (MTLR) showed remarkable predictability for almost all the datasets. We believe this outstanding performance presents a unique opportunity for a wider use of MTLR for survival risk factors. For translational clinicians and scientists, we hope our findings provide practical guidance for benchmark studies of survival models, as well as highlight potential areas of research interest.

Prediction Method of Rock Uniaxial Compressive Strength Based on Feature Optimization and SSA-XGBoost

In order to establish an optimal model for reasonably predicting the uniaxial compressive strength (UCS) of rocks, a method based on feature optimization and SSA-XGBoost was proposed. Firstly, the UCS predictor system of rocks, considering petrographic and physical parameters, was determined based on the systematic discussion of the factors affecting the UCS of rocks. Then, a feature selection method combining the RReliefF algorithm and Pearson correlation coefficient was proposed to further determine the optional input features. The XGBoost algorithm was used to establish the prediction model for rock UCS. In the process of model training, the Sparrow Search Algorithm (SSA) was used to optimize the hyperparameters. Finally, model evaluation was carried out to test the performance of the UCS prediction model. The method was applied and validated in a granitic tunnel. The results show that the proposed UCS prediction model can effectively predict the UCS of granitic rocks. Compared with simply adopting petrographic or physical parameters as the input features of the model, the UCS predictor considering petrographic and physical characteristics can improve the generalization ability of the SSA-XGBoost UCS prediction model effectively. The prediction method proposed in this study is reasonable and can provide some reference for establishing a universal method for accurately and quickly predicting the UCS of rocks.

Combining 2.5D deep learning and conventional features in a joint model for the early detection of sICH expansion.

The study aims to investigate the potential of training efficient deep learning models by using 2.5D (2.5-Dimension) masks of sICH. Furthermore, it intends to evaluate and compare the predictive performance of a joint model incorporating four types of features with standalone 2.5D deep learning, radiomics, radiology, and clinical models for early expansion in sICH. A total of 254 sICH patients were enrolled retrospectively and divided into two groups according to whether the hematoma was enlarged or not. The 2.5D mask of sICH is constructed with the maximum axial, coronal and sagittal planes of the hematoma, which is used to train the deep learning model and extract deep learning features. Predictive models were built on clinic, radiology, radiomics and deep learning features separately and four type features jointly. The diagnostic performance of each model was measured using the receiver operating characteristic curve (AUC), Accuracy, Recall, F1 and decision curve analysis (DCA). The AUCs of the clinic model, radiology model, radiomics model, deep learning model, joint model, and nomogram model on the train set (training and Cross-validation) were 0.639, 0.682, 0.859, 0.807, 0.939, and 0.942, respectively, while the AUCs on the test set (external validation) were 0.680, 0.758, 0.802, 0.857, 0.929, and 0.926. Decision curve analysis showed that the joint model was superior to the other models and demonstrated good consistency between the predicted probability of early hematoma expansion and the actual occurrence probability. Our study demonstrates that the joint model is a more efficient and robust prediction model, as verified by multicenter data. This finding highlights the potential clinical utility of a multifactorial prediction model that integrates various data sources for prognostication in patients with intracerebral hemorrhage. The Critical Relevance Statement: Combining 2.5D deep learning features with clinic features, radiology markers, and radiomics signatures to establish a joint model enabling physicians to conduct better-individualized assessments the risk of early expansion of sICH.

Development, assessment and validation of a novel nomogram model for predicting stroke mimics in stroke center：A single-center observational study

BackgroundEarly recognition and prediction of stroke mimics (SM) can avoid inappropriate recanalization therapy and delay in the management of SM etiology. The purpose of this study is to screen the predictors for SM and develop a novel predictive nomogram model for predicting SM. Meanwhile, the diagnostic performance of the nomogram model was evaluated and validated. The diagnostic efficacy of the nomogram model was also compared with four other SM structured scales. MethodsThe clinical data of eligible patients were retrospectively enrolled as training datasets from January 2020 to December 2021; and the clinical data of eligible patients were prospectively enrolled as validation datasets from February to December 2022 in stroke center, Shengjing hospital, respectively. Univariate analysis and Lasso regression were used to select the optimal predictors for the training set, and a nomogram model was constructed by multivariate logistics regression, predictive scoring based on nomogram model is performed for each subject suffering from suspected acute ischemic stroke. Area under the curve (AUC), Hosmer-Lemeshow goodness-of-fit test, Calibration curve, decision curve analysis (DCA), clinical impact curve (CIC) analysis and bootstrap sampling were performed to assess and validate the predictive performance and clinical utility of the nomogram model, and the DeLong test was used to compare the overall diagnostic performance of the nomogram model with the other four structured SM scales. The Delong test was also conducted to assess the external reliability of the SM nomogram model by comparing the predictive diagnostic performance of the validation set with the training set. Additionally, the Calibration curve was utilized to evaluate the diagnostic calibration capability of the SM nomogram model in the validation set. Results703 eligible patients (68 with SM, accounting for 9.7 %) were assigned to the training set, while 301 patients (26 with SM, accounting for 8.6 %) were assigned to the validation set. A nomogram model was then developed using these six parameters (SBP, history of epilepsy, isolated dizziness, isolated sensory impairment, headache, and absence of speech impairment symptoms), a dynamic web-based version of the nomogram was subsequently created. Comparing with four other scales, the nomogram model showed the highest overall diagnostic performance (AUC = 0.929, 95%CI = 0.908–0.947). The Hosmer-Lemeshow goodness-of-fit test was conducted to assess the agreement between the predicted SM values from the model and the observed SM values. The results of the test indicated a favorable consistency (χ2 = 9.299, P = 0.3177) between the predicted and observed SM. The results obtained from the analysis of the Calibration curve, DCA curve, and CIC analysis suggested that the nomogram possesses a favorable predictive capacity and superior clinical usefulness. Furthermore, the external validation demonstrated that there is no significant difference in the overall predictive diagnostic performance between the validation set and training set (0.929 vs 0.910, P > 0.05), thereby confirming the favorable stability of the nomogram model. ConclusionOur study firstly proposed a nomogram prediction approach based on the clinical features of SM, which could effectively predict the occurrence of SM. The utilization of the nomogram in stroke center proves advantageous for the identification and evaluation of SM, thereby enhancing diagnostic decision-making and strategies employed for suspected acute stroke patients.

Comparative analysis of some linear predictive models in the presence of multicollinearity

This paper investigates the comparative performance of some linear predictive models in the presence of multicollinearity. By examining the efficacy of Ordinary Least Squares (OLS), Ridge Regression, Lasso Regression, and Elastic Net Regression, this study aimed to figure out the best method for building robust and interpretable models under such conditions. The research explores how these models address multicollinearity, focusing on coefficient stability, prediction accuracy, and variable selection. Through a rigorous analysis of simulated and real-world datasets, the study shows the strengths and weaknesses of each model, providing valuable insights for researchers and practitioners looking to mitigate the challenges posed by multicollinearity in selecting the most proper method for regression modeling. This will lead to the creation of a model with increased interpretability of the relationships between variables, less variance, and more dependable coefficient estimations.

Mapping the soil C:N ratio at the European scale by combining multi-year Sentinel radar and optical data via cloud computing

Spatial information on the soil carbon-to-nitrogen (C:N) ratio is essential for sustainable soil use and management. The unprecedented availability of Sentinel optical and radar data on cloud computing platforms, such as the Google Earth Engine (GEE), has created new possibilities for developing soil prediction models from the local scale to the planetary scale. However, there is a paucity of literature on the effects of Sentinel sensor selection and integration and radar data utilization strategies on mapping the C:N ratio. In this study, we explored the use of multiyear Sentinel-1 radar and Sentinel-2 optical data obtained from the GEE platform combined with the digital soil mapping (DSM) technique to map the soil C:N ratio at the European scale. The performance of soil prediction models, which were constructed using two modeling techniques (random forest and support vector machine), derived under multiple scenarios based on optical, radar and commonly used auxiliary data (climatic and topographic variables) combined with the LUCAS 2018 soil dataset, was evaluated by a cross-validation technique. The results showed that the modeling performance varied with the selection and integration of Sentinel observations, as well as the configuration of the radar data. Models based on single polarization performed the worst across all scenarios related to Sentinel-1, with cross-polarization performing better than copolarization. Models that utilized Sentinel-1 data from ascending orbits outperformed those that utilized data from descending orbits. The application of Sentinel-1 backscatter information derived from different orbits and polarization modes resulted in improved prediction accuracy. Our study also demonstrated the potential of integrating multiyear Sentinel satellite data via the GEE to map the continental-scale C:N ratio. The model based on Sentinel-1 data outperformed the one built on Sentinel-2 data, whereas combining Sentinel-2 optical data with Sentinel-1 radar data led to more accurate predictions. The variable importance results indicated that optical data and backscattering information from Sentinel observations are the most important groups of variables for soil C:N ratio mapping compared to the other variable groups (terrain and climate data). The digital soil maps generated under the different scenarios exhibited detailed patterns with significant spatial variation, with similar overall trends but slightly different details.

Non-alcoholic fatty liver disease: A new predictor of recurrent ischemic stroke and transient ischemic attack in patients with carotid atherosclerosis

PurposeNon-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease. This study focused on assessing the predictive significance of NAFLD for recurrent stroke and transient ischemic attack (TIA) risk to determine the value of NAFLD. MethodThis study included 742 participants (mean age: 64.26 ± 9.42 years, 497 males) with carotid atherosclerosis who underwent carotid CT angiography (CTA) between January 2013 and December 2021 in this retrospective study. NAFLD was diagnosed by non-enhanced abdominal CT. The clinical endpoint was a recurrent ischemic stroke or TIA. Cox proportional hazards and Kaplan-Meier analysis assessed whether NAFLD was associated with the endpoint. We accessed the predictive values of NAFLD, clinical, plaque characteristics, and combined model by the C statistics. The predictive performance of the combined model was assessed by receiver operating characteristic curve (ROC) analysis. ResultsA total of 742 participants (mean age: 64.26 ± 9.42 years, 497 males) were included. During 2.9 years of follow-up (interquartile range, 2.1–3.9), 166 patients reached the clinical endpoint. Multivariable cox analyses showed NAFLD was associated with recurrent stroke or TIA in all groups (all P<0.05). Patients with NAFLD had a lower event-free survival (EFS) rate than those without NAFLD (P<0.05). The combined model, including NAFLD, clinical data and plaque features, showed the best performance in predicting the clinical endpoint (AUC=0.79). ConclusionsNAFLD contributes to the prediction of recurrent ischemic stroke or TIA. NAFLD may be a novel imaging marker that offers a new perspective on preventing cardiovascular disease in the clinic.

Optimizing Deep Learning Models with Improved BWO for TEC Prediction.

The prediction of total ionospheric electron content (TEC) is of great significance for space weather monitoring and wireless communication. Recently, deep learning models have become increasingly popular in TEC prediction. However, these deep learning models usually contain a large number of hyperparameters. Finding the optimal hyperparameters (also known as hyperparameter optimization) is currently a great challenge, directly affecting the predictive performance of the deep learning models. The Beluga Whale Optimization (BWO) algorithm is a swarm intelligence optimization algorithm that can be used to optimize hyperparameters of deep learning models. However, it is easy to fall into local minima. This paper analyzed the drawbacks of BWO and proposed an improved BWO algorithm, named FAMBWO (Firefly Assisted Multi-strategy Beluga Whale Optimization). Our proposed FAMBWO was compared with 11 state-of-the-art swarm intelligence optimization algorithms on 30 benchmark functions, and the results showed that our improved algorithm had faster convergence speed and better solutions on almost all benchmark functions. Then we proposed an automated machine learning framework FAMBWO-MA-BiLSTM for TEC prediction, where MA-BiLSTM is for TEC prediction and FAMBWO for hyperparameters optimization. We compared it with grid search, random search, Bayesian optimization algorithm and beluga whale optimization algorithm. Results showed that the MA-BiLSTM model optimized by FAMBWO is significantly better than the MA-BiLSTM model optimized by grid search, random search, Bayesian optimization algorithm, and BWO.

Enhancing composite cathode manufacturing with machine learning for polymer electrolyte solid-state batteries

Solid-state batteries (SSBs) represent a pivotal advancement in battery technology, poised to surpass lithium-ion batteries and drive the electrification of mobility. However, achieving cost-effective, scalable, and sustainable fabrication processes for SSB components remains a challenge. This study integrates machine learning (ML) techniques to optimize manufacturing processes of cathodes for polymer electrolyte based SSBs. The findings reveal strong predictive performance of regression models for active material loading, with support vector machine emerging as the top performer. Multiclassification models exhibit satisfactory precision, particularly in categorizing ideal electrode samples. Analysis of principal component and correlation circles highlight viscosity and wet thickness as critical variables for mixing and coating, respectively. Despite promising metrics, dataset imbalance and size limit model robustness. Further dataset augmentation is recommended before deployment in production. ML techniques offer promise in advancing battery manufacturing, paving the way for enhanced SSB performance and broader application across battery components.

An Empirical Study of the Impact of Class Overlap on the Performance and Interpretability of Cross-Version Defect Prediction

The class overlap problem refers to instances from different categories heavily overlapping in the feature space. This issue is one of the challenges in improving the performance of software defect prediction (SDP). Currently, the studies on the impact of class overlap on SDP mainly focused on within-project defect prediction and cross-project defect prediction. Moreover, the existing class overlap instances cleaning methods are not suitable for cross-version defect prediction. In this paper, we propose a class overlap instances cleaning method based on the Ratio of K-nearest neighbors with the Same Label (RKSL). This method removes instances with the abnormal neighbor ratio in the training set. Based on the RKSL method, we investigate the impact of class overlap on the performance and interpretability of the cross-version defect prediction model. The experiment results show that class overlap can affect the performance of cross-version defect prediction models significantly. The RKSL method can handle the class overlap problem in defect datasets, but it may impact the interpretability of models. Through the analysis of feature changes, we consider that class overlap instances cleaning can assist models in identifying more important features.

Novel hippocampus-centered methodology for informative instance selection in Alzheimer's disease data

The quantity and quality of a dataset play a crucial role in the performance of prediction models. Increasing the amount of data increases the computational requirements and can introduce negligible variations, outliers, and noise. These significantly impact the model performance. Thus, instance selection techniques are crucial for building prediction models with informative data, reducing the dataset size, improving performance, and minimizing computational costs. This study proposed a novel methodology for identifying the most informative two-dimensional slices derived from magnetic resonance imaging (MRI) to study Alzheimer's disease. The efficacy of our methodology was attributable to a hippocampus-centered analysis using data from multiple atlases. The methodology was evaluated by constructing convolutional neural networks to identify Alzheimer's disease, using a consolidated dataset constructed from three standard datasets: Alzheimer's Disease Neuroimaging Initiative, Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing, and Open Access Series of Imaging Studies. The proposed methodology demonstrated a commendable subject-level classification accuracy of approximately (95.00%) when distinguishing between normal cognition and Alzheimer's.

NecroGlobalGCN: Integrating micronecrosis information in HCC prognosis prediction via graph convolutional neural networks

Background and ObjectiveHepatocellular carcinoma (HCC) ranks fourth in cancer mortality, underscoring the importance of accurate prognostic predictions to improve postoperative survival rates in patients. Although micronecrosis has been shown to have high prognostic value in HCC, its application in clinical prognosis prediction requires specialized knowledge and complex calculations, which poses challenges for clinicians. It would be of interest to develop a model to help clinicians make full use of micronecrosis to assess patient survival. MethodsTo address these challenges, we propose a HCC prognosis prediction model that integrates pathological micronecrosis information through Graph Convolutional Neural Networks (GCN). This approach enables GCN to utilize micronecrosis, which has been shown to be highly correlated with prognosis, thereby significantly enhancing prognostic stratification quality. We developed our model using 3622 slides from 752 patients with primary HCC from the FAH-ZJUMS dataset and conducted internal and external validations on the FAH-ZJUMS and TCGA-LIHC datasets, respectively. ResultsOur method outperformed the baseline by 8.18% in internal validation and 9.02% in external validations. Overall, this paper presents a deep learning research paradigm that integrates HCC micronecrosis, enhancing both the accuracy and interpretability of prognostic predictions, with potential applicability to other pathological prognostic markers. ConclusionsThis study proposes a composite GCN prognostic model that integrates information on HCC micronecrosis, collecting large dataset of HCC histopathological images. This approach could assist clinicians in analyzing HCC patient survival and precisely locating and visualizing necrotic tissues that affect prognosis. Following the research paradigm outlined in this paper, other prognostic biomarker integration models with GCN could be developed, significantly enhancing the predictive performance and interpretability of prognostic model.

Predicting sleep based on physical activity, light exposure, and Heart rate variability data using wearable devices

Objective We aimed to improve the performance of sleep prediction algorithms by increasing the data amount, adding variables reflecting psychological state, and adjusting the data length. Materials and Methods We used ActiGraph GT3X+® and Galaxy Watch Active2™ to collect physical activity and light exposure data. We collected heart rate variability (HRV) data with the Galaxy Watch. We evaluated the performance of sleep prediction algorithms based on different data sources (wearable devices only, sleep diary only, or both), data lengths (1, 2, or 3 days), and analysis methods. We defined the target outcome, ‘good sleep’, as ≥90% sleep efficiency. Results Among 278 participants who denied having sleep disturbance, we used data including 2136 total days and nights from 230 participants. The performance of the sleep prediction algorithms improved with an increased amount of data and added HRV data. The model with the best performance was the extreme gradient boosting model; XGBoost, using both sources combined data with HRV, and 2-day data (accuracy=.85, area under the curve =.80). Conclusions The results show that the performance of the sleep prediction models improved by increasing the data amount and adding HRV data. Further studies targeting insomnia patients and applied researches on non-pharmacological insomnia treatment are needed.

Comparison of different MRI-based unsupervised segmentation algorithms in predicting sarcopenia

PurposeTo compare the performance of MRI-based Gaussian mixture model (GMM), K-means clustering, and Otsu unsupervised algorithms in predicting sarcopenia and to develop a combined model by integrating clinical indicators. MethodsRetrospective analysis was conducted on clinical and lumbar MRI data from 118 patients diagnosed with sarcopenia and 222 patients without the sarcopenia. All patients were randomly divided into training and validation groups in a 7:3 ratio. Regions of interest (ROI), specifically the paravertebral muscles at the L3/4 intervertebral disc level, were delineated on axial T2-weighted images (T2WI). The Gaussian mixture model (GMM), K-means clustering, and Otsu’s thresholding algorithms were employed to automatically segment muscle and adipose tissues at both the cohort and case levels. Subsequently, the mean signal intensity, volumes, and percentages of these tissues were calculated and compared. Logistic regression analyses were conducted to construct models and identify independent predictors of sarcopenia. An combined model was developed by combining the optimal magnetic resonance imaging (MRI) model and clinical predictors. The performance of the constructed model was assessed using receiver operating characteristic (ROC) curve analysis. ResultsAge, BMI, and serum albumin were identified as independent clinical predictors of sarcopenia. The cohort-level GMM demonstrated the best predictive performance both in the training group (AUC=0.840) and validation group (AUC=0.800), while the predictive performance of the other models was lower than that of the clinical model both in the training and validation groups. After combining the cohort-level GMM with the independent clinical predictors, the AUC of the training and validation groups increased to 0.871 and 0.867, respectively. ConclusionThe cohort-level GMM shows potential in predicting sarcopenia, and the incorporation of independent clinical predictors further increased the performance.

A centralized power prediction method for large-scale wind power clusters based on dynamic graph neural network

The uncertainty of the wind process leads to the randomness of its regional propagation, and the spatial correlation between nearby wind farms also shows a dynamic change tendency under the influence of wind direction. To consider the influence of dynamic spatial correlation on wind power prediction modeling, a short-term wind power prediction method based on a dynamic graph neural network is proposed. First, the topology graph of the wind farm cluster was established to represent the correlation of wind farms based on graph theory. Then, a dynamic spatiotemporal graph neural network was constructed to adapt graph topology by node embedding, which can explore the changing characteristics of spatial correlation among wind farms. Finally, we proposed a decoupling error model that can quantify the proportion of errors caused by the modeling process, which can assist in evaluating the performance of predictive models. We conducted experiments using data provided by the wind farm cluster in Inner Mongolia, and the average normalized Root Mean Square Error for the 12 wind farms was 0.1424, which verified the effectiveness of the proposed model.

A comparative study of ensemble methods and multi-output classifiers for predictive maintenance of hydraulic systems

The maintenance and upkeep of hydraulic systems play a pivotal role in various industrial applications, from manufacturing processes to heavy machinery operations. Traditional strategies often face significant limitations, particularly concerning intervention time and the costs associated with delays between fault occurrence and maintenance. Over the past decades, proactive strategies have emerged with promising potential, primarily due to their predictive capabilities. These proactive approaches aim to anticipate faults and maintenance needs, thereby mitigating costs and operational disruptions associated with reactive approaches. This study explores the applicability of single output, ensemble methods, and the integration of multi-output classifiers in hydraulic predictive maintenance problems with relatively limited data. First, data is analyzed using Pearson correlation coefficients, and then feature extraction is conducted using recursive feature extraction, aiming to optimize the performance of predictive models, particularly Random Forest and CatBoost. Results show that the stacking ensemble method, incorporating LightGBM, XGBoost, CatBoost, and Random Forest, yields the most notable improvement, achieving a final accuracy of 98.63%. The results obtained in this study show satisfying performances for single-output models, ensemble methods, and multi-output models in predicting the health of hydraulic systems. Moreover, combining single output, ensemble methods, and the integration of multi-output classifiers has created a relatively reliable and robust predictive maintenance system.

Lymphocyte to high density lipoprotein ratio can predict the short-term prognosis of hepatitis B virus-related acute-on-chronic liver failure patients

BackgroundAcute-on-chronic liver failure (ACLF) is a syndrome characterized by systemic inflammation, leading to high short-term mortality. The lymphocyte to high-density lipoprotein ratio (LHR) has been introduced as a novel marker of inflammation. However, its role as a prognostic inflammatory biomarker in the context of hepatitis B virus-related acute-on-chronic liver failure (HBV-ACLF) has received limited attention. MethodsWe retrospectively included 272 patients with HBV-ACLF who met the definition of APALC. Data on clinical features and laboratory tests were collected from medical records within 24 h. Logistic regression was used to identify risk factors for poor short-term prognosis, and LHR-based prediction (LHRB) models were constructed based on risk factors. Furthermore, the accuracy of the LHRB model was validated through rigorous testing. ResultsIn the survival and death groups, there were statistical differences in their CTP, MELD, MELD-Na, COSSH-ACLF II scores, and LHR. Multivariate logistic regression identified seven predictors significantly associated with 28-day mortality. Furthermore, statistically significant differences in short-term mortality and certain clinical laboratory tests for poor prognosis were observed between the high and low LHR groups. To assess the predictive performance of various models in terms of short-term mortality, the area under the receiver operating characteristic curve (AUROC) was calculated. The AUROC values for the CTP, MELD, MELD-Na, COSSH-ACLF II, and LHRB models were found to be 0.725, 0.788, 0.772, 0.871, and 0.877, respectively. The results in the validation group were similar to those in the training group, and the validation results suggested excellent performance of the LHRB model. ConclusionLHR levels have the potential to serve as indicators for the prognosis of HBV-ACLF. Additionally, the recently developed LHRB model offers an accessible risk assessment tool.

Machine-Learning Predictions of Critical Temperatures from Chemical Compositions of Superconductors.

In the quest for advanced superconducting materials, the accurate prediction of critical temperatures (Tc) poses a formidable challenge, largely due to the complex interdependencies between superconducting properties and the chemical and structural characteristics of a given material. To address this challenges, we have developed a machine-learning framework that aims to elucidate these complicated and hitherto poorly understood structure-property and property-property relationships. This study introduces a novel machine-learning-based workflow, termed the Gradient Boosted Feature Selection (GBFS), which has been tailored to predict Tc for superconductors by employing a distributed gradient-boosting framework. This approach integrates exploratory data analyses, statistical evaluations, and multicollinearity reduction techniques to select highly relevant features from a high-dimensional feature space, derived solely from the chemical composition of materials. Our methodology was rigorously tested on a data set comprising approximately 16,400 chemical compounds with around 12,000 unique chemical compositions. The GBFS workflow enabled the development of a classification model that distinguishes compositions likely to exhibit Tc values greater than 10 K. This model achieved a weighted average F1-score of 0.912, an AUC-ROC of 0.986, and an average precision score of 0.919. Additionally, the GBFS workflow underpinned a regression model that predicted Tc values with an R2 of 0.945, an MAE of 3.54 K, and an RMSE of 6.57 K on a test set obtained via random splitting. Further exploration was conducted through out-of-sample Tc predictions, particularly those exceeding the liquid nitrogen temperature, and out-of-distribution predictions for (Ca1-xLax)FeAs2 based on varying lanthanum content. The outcome of our study underscores the significance of systematic feature analysis and selection in enhancing predictive model performance, offering various advantages over models that rely primarily on algorithmic complexity. This research not only advances the field of superconductivity but also sets a precedent for the application of machine learning in materials science.