High Prediction Accuracy Research Articles

Estimation of Directional Wave Spectra with Motion Data of Floating Structure Using Complex-Valued Neural Networks

This study presents the development and validation of a complex-valued neural network model to predict wave conditions—such as wave direction, height, and period—surrounding floating structures. Accurate wave predictions are crucial for optimizing the design, operation, and maintenance of offshore platforms and floating offshore wind turbines, particularly in the context of digital twins. The proposed methodology leverages motion data obtained through numerical simulations of a floating structure to train the prediction model, enabling it to predict both the amplitude and phase information of the surrounding waves. The model successfully addresses the challenges of representing wave direction data in polar coordinates and capturing phase differences between motion components, which are difficult for traditional real-valued neural networks. The performance of the model was validated through various test cases, with the maximum prediction error found to be less than 10% and most predictions showing an error of less than 5%. Wave direction predictions demonstrated high accuracy, with errors consistently below 2%. While the model was trained using pseudo-measurement data, the results suggest that high-accuracy predictions can be achieved using real-world measurement data. This work contributes to enhancing wave prediction models for floating structures and is expected to improve the safety, performance, and long-term stability of such systems.

Oakland score to identify low-risk patients with lower gastrointestinal bleeding performs well among emergency department patients

BackgroundThe Oakland Score predicts risk of 30-day adverse events among hospitalized patients with lower gastrointestinal bleeding (LGIB) possibly identifying patients who may be safe for discharge. The Oakland Score has not been studied among emergency department (ED) patients with LGIB. The Oakland Score composite outcome includes re-bleeding, defined as additional blood transfusion requirements and/or a further decrease in hematocrit (Hct) >/= 20% after 24 h in clinical stability; red blood cell transfusion; therapeutic intervention to control bleeding, including surgery, mesenteric embolization, or endoscopic hemostasis; in-hospital death, all cause; and re-admission with further LGIB within 28 days. Prediction variables include age, sex, previous LGIB admission, systolic blood pressure, heart rate, and hemoglobin concentration, and scores range from 0 to 35 points, with higher scores indicating greater risk.MethodsRetrospective cohort study of adult (≥ 18 years old) patients with a primary ED diagnosis of LGIB across 21 EDs from March 1st, 2018, through March 1st, 2020. We excluded patients who were more likely to have upper gastrointestinal bleeding (esophago-gastroduodenoscopy without LGIB evaluation), patients who left against medical advice or prior to ED provider evaluation, ED patients without active health plan membership, and patients with incomplete Oakland Score variables. We assessed predictive accuracy by reporting the area under the receiver operator curve (AUROC) and sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios at multiple clinically relevant thresholds.ResultsWe identified 8,283 patients with LGIB, 52% were female, mean age was 68, 49% were non-White, and 27% had an adverse event. The AUROC for predicting an adverse event was 0.85 (95% CI 0.84–0.86). There were 1,358 patients with an Oakland Score of </=8; 4.9% had an adverse event, and sensitivity of the Oakland Score at this threshold was 97% (95% CI 96%−98%).ConclusionThe Oakland Score had high predictive accuracy among ED patients with LGIB. Prospective evaluation is needed to understand if the risk score could augment ED decision-making and improve outcomes and resource utilization.

Prognostic evaluation of non-muscle invasive bladder cancer with P-CRP and its nomogram

PurposeTo investigate the impact of the product of preoperative platelet count and C-reactive protein (P-CRP) on the postoperative prognosis of patients with non-muscle invasive bladder cancer (NMIBC), and to construct a Nomogram to predict the recurrence-free survival (RFS) of NMIBC patients based on pathological data.MethodsA retrospective analysis was conducted on the clinical data of 164 NMIBC patients who underwent transurethral resection of bladder tumors (TURBT) at the Second Affiliated Hospital of University of South China from January 2013 to December 2019. The endpoint of the study was the RFS. Kaplan-Meier (KM) method and Cox regression were used for analysis to identify independent factors affecting RFS. Then, the Nomogram was used to visualize the results of the multivariate analysis that were statistically significant and related to the RFS of NMIBC patients. Finally, the predictive ability of the model was evaluated using the concordance index (C-index) and calibration curves.ResultsBefore the end of the follow-up, the RFS was 88.3% at 1 year, 75.5% at 2 years, and 58.5% at 3 years. KM curves showed that P-CRP (HR=0.357, 95% CI: 0.204-0.625, P&amp;lt;0.001), number of tumors (HR=2.658, 95% CI: 1.572-4.494, P&amp;lt;0.001), tumor size (HR=2.271, 95% CI: 1.377-3.745, P=0.001), T stage of the tumor (HR=2.026, 95% CI: 1.233-3.329, P=0.005), and tumor G grade (G2: HR=1.615, 95% CI: 0.48-5.433, G3: HR=3.361, 95% CI: 1.022-11.054) were independent factors affecting the RFS of NMIBC patients after TURBT. The Nomogram could estimate the risk of tumor recurrence at 1, 2, and 3 years postoperatively. The Nomogram model incorporating P-CRP parameters had a higher predictive accuracy than the classic model that only included EORTC risk group parameters.ConclusionPreoperative P-CRP has a certain impact on the RFS of NMIBC patients after TURBT. The Nomogram incorporating P-CRP, number of tumors, tumor size, T stage, and tumor pathological grading can better predict the postoperative recurrence risk of NMIBC patients.

Machine Learning-Based Prediction of Drug Solubility in Lipidic Environments: The Sol_ME Tool for Optimizing Lipid-Based Formulations with a Preliminary Apalutamide Case Study.

Lipid-based formulations are essential for enhancing drug solubility and bioavailability, yet selecting optimal lipid excipients for specific drugs remains challenging. This study introduces Sol_ME, a machine learning-based model designed to predict drug solubility in lipidic environments, thereby streamlining the formulation process.The Sol_ME model uses PubChem® fingerprints, focusing on solubility correlations with lipid excipients, minimizing reliance on traditional parameters like LogP and molecular weight. The model was trained on a dataset of 1,379 drug-solvent entries and applied to the formulation of Apalutamide, a BCS Class II drug. Experimental validation was performed with 35 drug-solvent combinations to assess the accuracy of predicted solubilities.Sol_ME achieved a high predictive accuracy with a correlation coefficient of 0.998. The model successfully identified Cinnamon oil as the optimal excipient for Apalutamide, further refining the formulation with Vanillin. This reduced formulation volume by 75%, enabling the development of a single-unit 240mg soft gelatin capsule. Experimental validation showed 80% alignment between predicted and actual solubilities.The Sol_ME model demonstrates significant potential to optimize lipid-based drug formulation, offering a data-driven approach that enhances efficiency. The success of Apalutamide formulation highlights its practical utility. Future work will expand the dataset and extend the model to solid lipid systems, broadening its application in drug delivery technologies.

Exploring the feasibility of near-infrared spectroscopy in analyzing and predicting the chemical and biochemical changes of chitin-amended soils

Chitin, a natural polymer, has been added to soil to enhance plant resistance to pathogens, yet its impact on soil’s chemical and biochemical properties remains unclear. Standard methods to measure these properties are often slow and labor-intensive. This study aimed to assess pH, total DNA, chitinase activity, and trpB gene DNA targeting for the identification of pathogen Streptomyces scabies in chitin-treated soil using both conventional wet chemical analysis methods and near-infrared (NIR) spectroscopy for rapid measurement. Five soil groups were treated with varying chitin levels (control, 0.2% and 2% fungus chitin, 0.2% and 2% Sigma chitin) and sampled every 9 days over 122 days. NIR reflectance spectra (900-1,685 nm) from the soil samples were acquired at each sampling time. Partial least squares regression models were developed using preprocessing methods like Kubelka-Munk, second derivative, and smoothed data. The second derivative model for 900-1,660 nm yielded high predictive accuracy with R² values of 0.915, 0.892, and 0.810 for pH, total DNA, and trpB gene DNA respectively. This study demonstrates the potential of NIR spectroscopy coupled with partial least squares algorithms as a rapid, cost-effective method for estimating soil biochemical properties.

Machine learning models for quantitatively prediction of toxicity in macrophages induced by metal oxide nanoparticles.

As nanotechnology advances, metal oxide nanoparticles (MeONPs) increasingly come into contact with humans. The inhaled MeONPs cannot be effectively cleared by cilia or lung mucus. In the last decade, potential immune toxicity arising from exposure to MeONPs has been extensively debated, as lung macrophage is the main pathway for cleaning inhaled exogenous particles. However, their toxicity on lung macrophages has rarely been quantitatively predicted in silico due to the complexity of responses in macrophages and the intricate properties of MeONPs. Here, machine learning (ML) methods were used to establish models for quantitatively predicting the toxicity of MeONPs in macrophages. A multidimensional dataset including 240 data points covering the lethality, biochemical behaviors, and physicochemical properties of 30 MeONPs was obtained. ML models based on different algorithms with high prediction accuracy were constructed by addressing the issue of class imbalance during the training process. The models were verified by 10-fold cross-validation and external validation. The best-performed model has an R2 of 0.85 and 0.90 in the 10-fold cross-validation and external test set, respectively; and Q2 of 0.88 and 0.90 in the 10-fold cross-validation and test set, respectively. Five parameters that impact toxicity were identified and the toxicity mechanisms were elucidated by ML analysis. The prediction results can be used to fill the data gap in the risk assessment of nanomaterials. The framework offers valuable insights for designing and utilizing safe nanoparticles, as well as aiding in decision-making processes aimed at protecting the environment and public health.

A dual-optimization building energy prediction framework based on improved dung beetle algorithm, variational mode decomposition and deep learning

Accurately predicting building energy consumption is essential for enhancing energy utilization efficiency in buildings. However, the inherent volatility and noise in building energy data, caused by diverse user behaviors and potential sensor errors, make significant challenges to energy consumption prediction. To address these issues, a dual-optimization framework (IDBO-VMD-IDBO-BiLSTM) for building energy consumption prediction, which incorporates improved dung beetle optimization algorithm (IDBO), variational mode decomposition (VMD), and bidirectional long short-term memory network (BiLSTM), was proposed. In this framework, IDBO firstly optimizes the VMD by adaptively determining its optimal parameters to decompose the original building energy consumption series into multiple intrinsic modal functions (IMFs) with smoother characteristics, thereby the effect of mitigating data noise. Then, each IMF component is predicted using the BiLSTM model, with IDBO selecting the optimal hyperparameters for BiLSTM. Finally, the individual predictions of each IMF are superimposed and reconstructed to yield the final predictions. To verify the framework’s effectiveness, real energy consumption data from an office building in Shanghai was collected and analyzed in a comprehensive comparison with seven other comparative models. Experimental results suggested that the proposed framework outperformed the comparative models in terms of root mean square error (RMSE), mean absolute percentage error (MAPE), mean absolute error (MAE), and coefficient of determination (R2), showing both high predictive accuracy and strong robustness. Therefore, the proposed framework can be an effective tool for predicting building energy consumption

Short communication: Genomic prediction based on unbiased estimation of the genomic relationship matrix in pigs.

The traditional genomic relationship matrix (GRM) has shown to be a biased estimation of true kinship, which can affect subsequent genetic analyses. In this study, we employed an unbiased kinship (UKin) estimation method within the genomic best linear unbiased prediction framework to evaluate its prediction performance on both a simulated dataset and a Large White pig dataset. The simulated dataset encompasses six traits, 900 quantitative trait loci, and 36000 single nucleotide polymorphisms (SNPs). Two scenarios (small effect genes; major genes and small effect genes) and three heritabilities (0.1, 0.3 and 0.5) were considered. The Large White pig dataset includes two traits, 3290 animals and 35172 SNPs. The prediction performance of the Ukin method was compared with several other GRM construction methods, including VanRaden1 and 2 methods, Goudet method, and the runs of homozygosity (ROH) method. In the simulated dataset, VanRaden2 method and the UKin+VanRaden1 method achieved relatively higher prediction accuracies, averaging 0.561 and 0.558 for the six traits, respectively. Apart from the ROH method, all methods demonstrated similar levels of unbiasedness, around 1.10. In the Large White pig dataset, the accuracy of two traits hovered around 0.780, and the unbiasedness around 0.99, again with the ROH method as an exception. This study underscores the potential of the unbiased kinship estimation method in animal breeding.

Dự báo giá cổ phiếu CATL đóng cửa và các chiến lược giao dịch sử dụng XGBoost và thuật toán tối ưu hóa Optuna

This study employs the XGBoost algorithm, enhanced through hyperparameter optimization with Optuna, to forecast the closing stock prices of CATL and develop effective trading strategies. The XGBoost model achieved superior performance, with an R² score of 0.9201, a Mean Absolute Error (MAE) of 0.1982, and a Mean Squared Error (MSE) of 0.0825, outperforming benchmark models such as ElasticNet and Decision Tree. The resulting trading strategy generated actionable buy, sell, and hold recommendations over a 30-day horizon, aiding investors in profit optimization. The findings highlight the high prediction accuracy and stability of the XGBoost model, particularly when optimized with Optuna, thereby enhancing decision-making efficiency in stock trading. Nonetheless, the study underscores the potential for further improvement through the integration of advanced deep learning models, alternative optimization techniques, or ensemble learning approaches. Future research could also explore incorporating unstructured data to refine forecasting accuracy and expand the applicability of trading strategies.

Genome-wide association mapping and genomic prediction analyses reveal the genetic architecture of grain yield and agronomic traits under drought and optimum conditions in maize

BackgroundDrought is a major abiotic stress in sub-Saharan Africa, impacting maize growth and development leading to severe yield loss. Drought tolerance is a complex trait regulated by multiple genes, making direct grain yield selection ineffective. To dissect the genetic architecture of grain yield and flowering traits under drought stress, a genome-wide association study (GWAS) was conducted on a panel of 236 maize lines testcrossed and evaluated under managed drought and optimal growing conditions in multiple environments using seven multi-locus GWAS models (mrMLM, FASTmrMLM, FASTmrEMMA, pLARmEB, pKWmEB, ISIS EM-BLASSO, and FARMCPU) from mrMLM and GAPIT R packages. Genomic prediction with RR-BLUP model was applied on BLUEs across locations under optimum and drought conditions.ResultsA total of 172 stable and reliable quantitative trait nucleotides (QTNs) were identified, of which 77 are associated with GY, AD, SD, ASI, PH, EH, EPO and EPP under drought and 95 are linked to GY, AD, SD, ASI, PH, EH, EPO and EPP under optimal conditions. Among these QTNs, 17 QTNs explained over 10% of the phenotypic variation (R2 ≥ 10%). Furthermore, 43 candidate genes were discovered and annotated. Two major candidate genes, Zm00001eb041070 closely associated with grain yield near peak QTN, qGY_DS1.1 (S1_216149215) and Zm00001eb364110 closely related to anthesis-silking interval near peak QTN, qASI_DS8.2 (S8_167256316) were identified, encoding AP2-EREBP transcription factor 60 and TCP-transcription factor 20, respectively under drought stress. Haplo-pheno analysis identified superior haplotypes for qGY_DS1.1 (S1_216149215) associated with the higher grain yield under drought stress. Genomic prediction revealed moderate to high prediction accuracies under optimum and drought conditions.ConclusionThe lines carrying superior haplotypes can be used as potential donors in improving grain yield under drought stress. Integration of genomic selection with GWAS results leads not only to an increase in the prediction accuracy but also to validate the function of the identified candidate genes as well increase in the accumulation of favorable alleles with minor and major effects in elite breeding lines. This study provides valuable insight into the genetic architecture of grain yield and secondary traits under drought stress.

Electro-Thermal Coupling Modeling and Heat Generation Decoupling Analysis of Semi-Solid-State Lithium-Ion Battery

Solid-state batteries are increasingly seen as the future of battery development due to their higher energy density and improved thermal stability. While all-solid-state batteries are gaining attention, semi-solid-state batteries, serving as a bridge between liquid batteries and all-solid-state ones, have not been extensively studied. This article delves into the electro-thermal characteristics of a commercial semi-solid-state battery through various experiments like HPPC, adiabatic calorimetry, and entropy heat coefficient analysis. It compares the internal resistance traits with those of a similar liquid lithium-ion pouch battery. By integrating a second-order RC equivalent circuit model with the Bernardi heat generation model, a comprehensive electro-thermal coupled model for semi-solid-state lithium-ion batteries is established. This model considers parameters such as temperature, state of charge, and charge/discharge rates to accurately predict terminal voltage, reversible and irreversible heat generation power, and temperature variations. The study demonstrates a high prediction accuracy with the maximum root mean square error of 0.043 V for terminal voltage, 0.67 W for heat generation power, and 0.56°C for temperature. Following the validation of the model, an analysis is conducted to dissect the different components contributing to battery heat generation. The insights gained from the characteristics analysis and electro-thermal model of semi-solid-state batteries can be valuable for enhancing the control strategies of battery management systems in the future.

Implications of the choroid plexus in Niemann-Pick disease Type C neuropathogenesis.

Niemann-Pick Disease Type C (NPC) is an ultra-rare disorder characterized by progressive psychiatric and neurologic manifestations, with late infantile, juvenile, and adolescent/adult presentations. We examined morphological properties of the choroid plexus, a protective blood-cerebrospinal fluid barrier, in NPC, and their relationship with neurodegeneration, clinical status, and circulatory markers. This study also determined whether choroid plexus morphology differentiates between NPC and more prevalent illnesses, schizophrenia (SZ) and bipolar disorder (BD), which have overlapping psychiatric symptoms with adolescent and adult-onset NPC and are associated with misdiagnosis. Patients with NPC were assessed using neuroimaging, clinical instruments, and plasma protein quantification focusing on inflammatory markers. Morphological properties (i.e., choroid plexus volumes) were compared between patients with NPC (n=17), SZ (n=20), BD (n=24), and healthy controls (HCs, n=106). Choroid plexus enlargement (p<0.05) and reduced thalamic volumes (p<0.05) were observed in NPC patients versus HCs and SZ or BD patients. A logistic regression model with choroid plexus and thalamic volumes as predictors yielded high prediction accuracy for NPC vs. HCs, NPC vs. SZ, and NPC vs. BD (area under the receiver operating characteristics curve [AUROC] of 1). Choroid plexus volumes were negatively correlated with left (p=0.009-0.012) and right (p=0.007-0.025) thalamic volumes, left (r=-0.69, p=0.003) and right (r=-0.71, p=0.002) crus I of the cerebellum, and greater severity on the NPC-Suspicion Index psychiatric subscale (ρ=0.72, p=0.042). Targeted protein expression quantification revealed differential expression of TGFA, HLA-DRA, TNFSF12, EGF, INFG, and IL-18 in NPC patients vs. HCs (p<0.05), with higher choroid plexus volumes correlating with IL-18 levels (ρ=0.71, p=0.047). The choroid plexus may play a critical role in NPC neuropathogenesis and serve as a novel biomarker for monitoring neurodegenerative and inflammatory processes in NPC.

Interpretable data-driven framework for life prediction of homogenous lead-free solder joints in ball grid array packages

Abstract Designing reliable electronic devices depends heavily on the reliability of solder joints. However, due to the ongoing advancement of novel lead-free solder materials, assessing their reliability for various applications, particularly in extreme conditions, often involves costly and time-consuming tests. Multiple factors, including board, package, assembly parameters, solder material properties, and process parameters, all affect the reliability of solder joints. Leveraging the available reliability test data as well as the simulated data, an interpretable data-driven framework is proposed by integrating linear mixed-effects (LME) models and artificial neural network (ANN) with sensitivity analysis to achieve accurate life predictions of solder materials and improve interpretability for understanding new lead-free materials. Multiple linear regression (MLR) and decision tree (DT) models have also been included for comparison. For thermal cycling (TC) tests, ANN, LME, and MLR achieve high prediction accuracy (R-squared values of 88.95%, 91.79%, and 90.07%, respectively), which are approximately 10%, 14%, and 13% higher than DT. Moreover, the interpretation derived from the random effects of LME and the sensitivity analysis of ANN, along with the insights from MLR and DT, successfully extract valuable information about the life of solder materials, which aligns with experimental testing results and empirical physics knowledge. For instance, it is found that the impact of aging duration on reliability depends on the configuration of the tested electronic package. Furthermore, as the temperature range of the TC test increases, the significance of the solder alloy material’s influence on reliability decreases. This reduction in relevance is associated with a transition to the failure mode under more extreme temperature conditions. Therefore, the interpretable data-driven framework not only models the complex relationships among influencing factors and solder joint reliability but also enables understanding the solder joint reliability. This framework can contribute to accelerating the design and adoption of new lead-free solder materials in various electronic applications and utilizing data-driven models to assess the reliability of electronic packages in future research.

A scoping review of automatic and semi-automatic MRI segmentation in human brain imaging.

AI-based segmentation techniques in brain MRI have revolutionized neuroimaging by enhancing the accuracy and efficiency of brain structure analysis. These techniques are pivotal for diagnosing neurodegenerative diseases, classifying psychiatric conditions, and predicting brain age. This scoping review synthesizes current methodologies, identifies key trends, and highlights gaps in the use of automatic and semi-automatic segmentation tools in brain MRI, particularly focusing on their application to healthy populations and clinical utility. A scoping review was conducted following Arksey and O'Malley's framework and PRISMA-ScR guidelines. A comprehensive search was performed across six databases for studies published between 2014 and 2024. Studies focused on AI-based brain segmentation in healthy populations, and patients with neurodegenerative diseases, and psychiatric disorders were included, while reviews, case series, and studies without human participants were excluded. Thirty-two studies were included, employing various segmentation tools and AI models such as convolutional neural networks for segmenting gray matter, white matter, cerebrospinal fluid, and pathological regions. FreeSurfer, which utilizes algorithmic techniques, are also commonly used for automated segmentation. AI models demonstrated high accuracy in brain age prediction, neurodegenerative disease classification, and psychiatric disorder subtyping. Longitudinal studies tracked disease progression, while multimodal approaches integrating MRI with fMRI and PET enhanced diagnostic precision. AI-based segmentation techniques provide scalable solutions for neuroimaging, advancing personalized brain health strategies and supporting early diagnosis of neurological and psychiatric conditions. However, challenges related to standardization, generalizability, and ethical considerations remain. The integration of AI tools and algorithm-based methods into clinical workflows can enhance diagnostic accuracy and efficiency, but greater focus on model interpretability, standardization of imaging protocols, and patient consent processes is needed to ensure responsible adoption in practice.

Identification of Prognosis Signature Based on cGAS-STING Pathway and Its Immunotherapeutic Significance in Lung Adenocarcinoma.

Lung adenocarcinoma (LUAD) is a leading cause of cancer-related deaths worldwide, and there is an urgent need to develop personalized prognostic models for effective treatment strategies. The cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathways has been confirmed to engage in multiple cancer progression, prognosis, and immunotherapy benefits. However, the prognostic significance and immunotherapy response of cGAS-STING pathway-associated genes (CSPAGs) in LUAD remain unclear. Herein, we aimed to establish a CSPAG-based prognostic signature for LUAD patients. A total of 139 CSPAGs derived from the GSEA website were enrolled for subsequent analysis. Univariate Cox regression analysis shows that 22 of 139 CSPAGs were associated with LUAD prognosis. Lasso analysis identified 6 CSPAGs (IFNE, NFKB2, POL3RG, TRAF2, TICAM1 and NLRC3) as the most significant prognostic CSPAGs with the best model efficacy. The CSPAG signature classified LUAD patients into low-risk (LR) and high-risk (HR) groups. Kaplan-Meier analysis demonstrated that patients in the LR group had significantly better overall survival (OS) than those in the HR group (p < 0.05 represents statistical significance), indicating the predictive power of the CSPAG signature in LUAD prognosis. The receiver operating characteristic (ROC) curve analysis showed that the area under the curve (AUC) values for the CSPAG signature were higher than those for other well-established predictive factors, suggesting that the CSPAG signature had a higher predictive efficacy. The CSPAG nomogram incorporating clinical factors such as age, TNM status and the CSPAG risk score accurately predicted the OS of LUAD patients at 1, 3, and 5 years, indicating its potential clinical application in LUAD prognosis. Furthermore, we investigated the expression pattern of the 6 signature CSPAGs in different LUAD subpopulations with distinct clinical features. The CSPAG risk score was increased in the immune-high groups, suggesting a positive correlation between immune infiltration degree and CSPAG risk score. There was a heterogenicity of somatic mutation landscape between the two groups. The LR group had a strong immune cell activity, and most immune checkpoints were significantly expressed in the LR group, implying that this group benefited from immune checkpoint blockade (ICB) therapy. In addition, we verified the high predictive accuracy of the CSPAG signature in the GSE31210 and GSE203360 datasets. Taken together, this study established a CSPAG-based prognostic signature for LUAD patients with high predictive efficacy and clinical relevance. The association between CSPAGs and immune infiltration, and ICB therapy response, highlights the potential of the CSPAG signature as a personalized treatment strategy for LUAD patients.

An improved theoretical model for eddy‐induced bubble breakup in turbulent flows

AbstractThis work aims on establishing an improved model for bubble breakup in turbulent flows. To achieve this goal, the impact of bubble shape on the critical size for distinguishing large and small eddies, cross‐sectional area of eddy‐bubble collision tube, energy increment breakage criterion, and force breakage criterion is deeply studied, and accurate formulas quantifying these effects are incorporated into the developed model. Moreover, a dimensionless number is proposed to overcome the limitation of the collision cube mechanism model which neglects the influence of bubble oscillation in the radial direction of the tube. The results indicate that the developed model exhibits high prediction accuracy with a mean absolute relative error (MARE) of 18.88% for the breakage frequency, and the predicted bubble size distribution aligns closely with experimental trends.

High-Accuracy Prediction of Mechanical Properties of Ni-Cr-Fe Alloys Using Machine Learning

Artificial intelligence-driven prediction models have emerged as powerful tools for estimating material properties with high accuracy, yet the preparation of training datasets often demands labor-intensive and time-consuming experimental procedures. Leveraging the Computational Materials Science (CMS) approach, this study utilizes phase transformation calculations and thermodynamic data to simulate the mechanical properties of Ni-Cr-Fe alloys. Using JMatPro software, mechanical properties (0.2% Proof Stress, Fracture Stress, and Young's Modulus) of 50 Ni-Cr-Fe alloy compositions were simulated across a temperature range of 540–920°C, generating a dataset of 1000 rows. This dataset was used to train an Artificial Neural Network (ANN) model, with 80% allocated for training and 20% for validation and testing. The trained AI model demonstrated robust predictive capabilities, achieving a 96,61% accuracy rate in forecasting material compositions with the desired thermo-physical properties at specific temperatures. To validate the model's reliability, predicted alloy compositions were re-simulated under identical conditions in JMatPro, confirming the high fidelity of the model's predictions. The results underscore the efficacy of Computational Materials Science (CMS)-generated datasets as a scalable and reliable source for training AI models in materials science. This study highlights the potential of integrating Computational Materials Science (CMS) and Machine Learning approaches to accelerate material design and development processes, delivering significant improvements in prediction speed and accuracy.

Dimensionless Machine Learning: Dimensional Analysis to Improve LSSVM and ANN models and predict bearing capacity of circular foundations

This research focuses on developing hybrid intelligence techniques to predict the bearing capacity of circular foundations using the most effective parameters. In this regard, a database involving 968 test circular foundations involving different rock properties, soil characteristics, and foundation radius has been prepared by laboratory tests for training and testing the new model presented in this study. For adequately considering various factors, the several parameters of depth (D) in m, density of soil (DS) in gr/cm3, internal angle of friction (IAF) in degree, cohesion of soil (CS) in kg/cm2, and foundation radius (FR) in m were considered as the input of the model and bearing capacity in kg/cm2 was considered as the target parameter. Three main strategies were addressed in this study. First, four well-known machine learning algorithms, Artificial Neural Network (ANN), Bagging Regressor (BR), Least Squares Support Vector Machine (LSSVM), and Gradient Boosting Regressor (GBR), were adopted to predict bearing capacity. Second, a novel and robust mathematical computation named dimensional analysis (DA) theorem was integrated with machine learning techniques to improve the accuracy and performance of the models by decreasing the number of inputs. Third, based on DA implementation, a rational mathematical formula using gene expression programming (GEP) was structured to anticipate bearing capacity. Furthermore, the sensitivity analysis process specified the impact of the five effective factors. The results of this study demonstrate the effectiveness of integrating DA with machine learning models to predict the BC of circular foundations. Among the developed models, the DA-LSSVM model showed superior performance, achieving an R² of 0.998 and 0.996 for training and testing, respectively, indicating high prediction accuracy. The results indicated that the IAF was the most sensitive factor with r = 0.709, while CS was the least sensitive with r = -0.087. A graphical user interface (GUI) app has been designed to apply the proposed models in this study conveniently. In the last step of this process, the GUI and the DA-based machine learning models were implemented by analyzing two examples. According to the findings, the GUI may be employed to make a reliable and speedy projection of the bearing capacity of circular foundations by considering a variety of input parameters.

A Study on the Energy Absorption Performance of Mine Grooved Conical Tube Energy Absorption Components

When rockbursts occur, hydraulic support is prone to impact failure, which leads to severe casualties and economic losses. To improve the performance of hydraulic support structures under impact loading, a grooved conical tube is designed as an energy absorption device to avoid hydraulic columns being destroyed. The performance of the grooved conical tube during deformation is studied using simulation, considering the wall thickness, cone angle and number of grooves. The equivalent axial load of the grooved conical tube component is derived by studying the energy dissipation path. And the grooved conical tube’s structure is optimized. The results show that the Y3-5-10 (cone angle: 3°; number of grooves: 5; wall thickness: 10 mm) grooved conical tube shows excellent performance among the twenty-seven types of structures. In addition, the equivalent axial load prediction formula for the grooved conical tube has a high prediction accuracy. Furthermore, after multi-objective optimization, the mean square error is decreased by 20.6%, and the effective energy absorption is increased by 6.0%, which is able to make the energy absorption process more stable. Compared with widely used corrugated square tubes, the effective deformation distance of the grooved conical tube is increased by 27.2%, and the effective energy absorption is increased by 37.1%. The grooved conical tube has advantages in its effective deformation distance and effective energy absorption. These results are expected to provide sufficient time for the opening of the support column’s relief valve and to enhance the impact resistance of the hydraulic support, which is highly important for the prevention of rockbursts.

Uncertainty Quantification in Shear Wave Velocity Predictions: Integrating Explainable Machine Learning and Bayesian Inference

The accurate prediction of shear wave velocity (Vs) is critical for earthquake engineering applications. However, the prediction is inevitably influenced by geotechnical variability and various sources of uncertainty. This paper investigates the effectiveness of integrating explainable machine learning (ML) model and Bayesian generalized linear model (GLM) to enhance both predictive accuracy and uncertainty quantification in Vs prediction. The study utilizes an Extreme Gradient Boosting (XGBoost) algorithm coupled with Shapley Additive Explanations (SHAPs) and partial dependency analysis to identify key geotechnical parameters influencing Vs predictions. Additionally, a Bayesian GLM is developed to explicitly account for uncertainties arising from geotechnical variability. The effectiveness and predictive performance of the proposed models were validated through comparison with real case scenarios. The results highlight the unique advantages of each model. The XGBoost model demonstrates good predictive performance, achieving high coefficient of determination (R2), index of agreement (IA), Kling–Gupta efficiency (KGE) values, and low error values while effectively explaining the impact of input parameters on Vs. In contrast, the Bayesian GLM provides probabilistic predictions with 95% credible intervals, capturing the uncertainty associated with the predictions. The integration of these two approaches creates a comprehensive framework that combines the strengths of high-accuracy ML predictions with the uncertainty quantification of Bayesian inference. This hybrid methodology offers a powerful and interpretable tool for Vs prediction, providing engineers with the confidence to make informed decisions.