Development Of Machine Learning Models Research Articles

MICROPHERRET: MICRObial PHEnotypic tRait ClassifieR using Machine lEarning Techniques

BackgroundIn recent years, there has been a rapid increase in the number of microbial genomes reconstructed through shotgun sequencing, and obtained by newly developed approaches including metagenomic binning and single-cell sequencing. However, our ability to functionally characterize these genomes by experimental assays is orders of magnitude less efficient. Consequently, there is a pressing need for the development of swift and automated strategies for the functional classification of microbial genomes.ResultsThe present work leverages a suite of supervised machine learning algorithms to establish a range of 86 metabolic and other ecological functions, such as methanotrophy and plastic degradation, starting from widely obtainable microbial genome annotations. Tests performed on independent datasets demonstrated robust performance across complete, fragmented, and incomplete genomes above a 70% completeness level for most of the considered functions. Application of the algorithms to the Biogas Microbiome database yielded predictions broadly consistent with current biological knowledge and correctly detecting functionally-related nuances of archaeal genomes. Finally, a case study focused on acetoclastic methanogenesis demonstrated how the developed machine learning models can be refined or expanded with models describing novel functions of interest.ConclusionsThe resulting tool, MICROPHERRET, incorporates a total of 86 models, one for each tested functional class, and can be applied to high-quality microbial genomes as well as to low-quality genomes derived from metagenomics and single-cell sequencing. MICROPHERRET can thus aid in understanding the functional role of newly generated genomes within their micro-ecological context.

Development of machine learning models for forecasting the strength of resilient modulus of subgrade soil: genetic and artificial neural network approaches

Accurately predicting the Modulus of Resilience (MR) of subgrade soils, which exhibit non-linear stress–strain behaviors, is crucial for effective soil assessment. Traditional laboratory techniques for determining MR are often costly and time-consuming. This study explores the efficacy of Genetic Programming (GEP), Multi-Expression Programming (MEP), and Artificial Neural Networks (ANN) in forecasting MR using 2813 data records while considering six key parameters. Several Statistical assessments were utilized to evaluate model accuracy. The results indicate that the GEP model consistently outperforms MEP and ANN models, demonstrating the lowest error metrics and highest correlation indices (R2). During training, the GEP model achieved an R2 value of 0.996, surpassing the MEP (R2 = 0.97) and ANN (R2 = 0.95) models. Sensitivity and SHAP (SHapley Additive exPlanations) analysis were also performed to gain insights into input parameter significance. Sensitivity analysis revealed that confining stress (21.6%) and dry density (26.89%) are the most influential parameters in predicting MR. SHAP analysis corroborated these findings, highlighting the critical impact of these parameters on model predictions. This study underscores the reliability of GEP as a robust tool for precise MR prediction in subgrade soil applications, providing valuable insights into model performance and parameter significance across various machine-learning (ML) approaches.

An XAI Framework for Predicting Wind Turbine Power under Rainy Conditions Developed Using CFD Simulations

Renewable energy sources are essential to address climate change, fossil fuel depletion, and stringent environmental regulations in the subsequent decades. Horizontal-axis wind turbines (HAWTs) are particularly suited to meet this demand. However, their efficiency is affected by environmental factors because they operate in open areas. Adverse weather conditions like rain reduce their aerodynamic performance. This study investigates wind turbine power prediction under rainy conditions by integrating Blade Element Momentum (BEM) theory with explainable artificial intelligence (XAI). The S809 airfoil’s aerodynamic characteristics, used in NREL wind turbines, were analyzed using ANSYS FLUENT and symbolic regression under varying rain intensities. Simulations at a Reynolds number (Re) of 1 × 106 were performed using the Discrete Phase Model (DPM) and k–ω SST turbulence model, with liquid water content (LWC) values of 0 (dry), 10, 25, and 39 g/m3. The lift and drag coefficients were calculated at various angles of attack for all the conditions. The results indicated that rain led to reduced lift and increased drag. The innovative aspect of this research is the development of machine learning models predicting changes in the airfoil coefficients under rain with an R2 value of 0.97. The proposed XAI framework models rain effects at a lower computational time, enabling efficient wind farm performance assessment in rainy conditions compared to conventional CFD simulations. It was found that a heavy rain LWC of 39 g/m3 could reduce power output by 5.7% to 7%. These findings highlight the impact of rain on aerodynamic performance and the importance of advanced predictive models for optimizing renewable energy generation.

Exploring the Potential of a Smart Ring to Predict Postoperative Pain Outcomes in Orthopedic Surgery Patients

Poor pain alleviation remains a problem following orthopedic surgery, leading to prolonged recovery time, increased morbidity, and prolonged opioid use after hospitalization. Wearable device data, collected during postsurgical recovery, may help ameliorate poor pain alleviation because a patient’s physiological state during the recovery process may be inferred from sensor data. In this study, we collected smart ring data from 37 inpatients following orthopedic surgery and developed machine learning models to predict if a patient had postsurgical poor pain alleviation. Machine learning models based on the smart ring data were able to predict if a patient had poor pain alleviation during their hospital stay with an accuracy of 70.0%, an F1-score of 0.769, and an area under the receiver operating characteristics curve of 0.762 on an independent test dataset. These values were similar to performance metrics from existing models that rely on static, preoperative patient factors. Our results provide preliminary evidence that wearable device data may help control pain after orthopedic surgery by incorporating real-time, objective estimates of a patient’s pain during recovery.

Development of machine learning models for fractional flow reserve prediction in angiographically intermediate coronary lesions.

Fractional flow reserve (FFR) represents the gold standard in guiding the decision to proceed or not with coronary revascularization of angiographically intermediate coronary lesion (AICL). Optical coherence tomography (OCT) allows to carefully characterize coronary plaque morphology and lumen dimensions. We sought to develop machine learning (ML) models based on clinical, angiographic and OCT variables for predicting FFR. Data from a multicenter, international, pooled analysis of individual patient's level data from published studies assessing FFR and OCT on the same target AICL were collected through a dedicated database to train (n = 351) and validate (n = 151) six two-class supervised ML models employing 25 clinical, angiographic and OCT variables. A total of 502 coronary lesions in 489 patients were included. The AUC of the six ML models ranged from 0.71 to 0.78, whereas the measured F1 score was from 0.70 to 0.75. The ML algorithms showed moderate sensitivity (range: 0.68-0.77) and specificity (range: 0.59-0.69) in detecting patients with a positive or negative FFR. In the sensitivity analysis, using 0.75 as FFR cut-off, we found a higher AUC (0.78-0.86) and a similar F1 score (range: 0.63-0.76). Specifically, the six ML models showed a higher specificity (0.71-0.84), with a similar sensitivity (0.58-0.80) with respect to 0.80 cut-off. ML algorithms derived from clinical, angiographic, and OCT parameters can identify patients with a positive or negative FFR.

Predicting metformin efficacy in improving insulin sensitivity among women with polycystic ovary syndrome and insulin resistance: a machine learning study

ObjectiveMetformin is clinically effective in treating polycystic ovary syndrome (PCOS) with insulin resistance (IR), while its efficacy varies among individuals. This study aims to develop a machine learning model to predict the efficacy of metformin in improving insulin sensitivity among women with PCOS and IR. MethodsThis is a retrospective analysis of a multicenter, randomized controlled trial involving 114 women diagnosed with PCOS and IR. All women received metformin treatment for 4 months. We incorporated 27 baseline clinical variables of the women into the construction of our machine learning model. We firstly compared four commonly used feature selection methods to screen valuable clinical variables. Then we used the valuable variables as inputs to evaluate the performance of five machine learning models, including k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (Xgboost), in predicting the efficacy of metformin. ResultsAmong the five machine learning models, SVM performed the best with an area under the receiver operating characteristic curve (AUC) of 0.781 (95% confidence interval [CI]: 0.772-0.791). The key predictive variables identified were homeostasis model assessment of insulin resistance (HOMA-IR), body mass index (BMI), and low-density lipoprotein cholesterol (LDL-C). ConclusionThe developed machine learning model could be applied to to predict the efficacy of metformin in improving insulin sensitivity among women with PCOS and IR. The result could help doctors evaluate the efficacy of metformin in advance, optimize treatment plans, and thereby enhance overall clinical outcomes.

Decoding pulsatile patterns of cerebrospinal fluid dynamics through enhancing interpretability in machine learning

Analyses of complex behaviors of Cerebrospinal Fluid (CSF) have become increasingly important in diseases diagnosis. The changes of the phase-contrast magnetic resonance imaging (PC-MRI) signal formed by the velocity of flowing CSF are represented as a set of velocity-encoded images or maps, which can be thought of as signal data in the context of medical imaging, enabling the evaluation of pulsatile patterns throughout a cardiac cycle. However, automatic segmentation of the CSF region in a PC-MRI image is challenging, and implementing an explained ML method using pulsatile data as a feature remains unexplored. This paper presents lightweight machine learning (ML) algorithms to perform CSF lumen segmentation in spinal, utilizing sets of velocity-encoded images or maps as a feature. The Dataset contains 57 PC-MRI slabs by 3T MRI scanner from control and idiopathic scoliosis participants are involved to collect data. The ML models are trained with 2176 time series images. Different cardiac periods image (frame) numbers of PC-MRIs are interpolated in the preprocessing step to align to features of equal size. The fivefold cross-validation procedure is used to estimate the success of the ML models. Additionally, the study focusses on enhancing the interpretability of the highest-accuracy eXtreme gradient boosting (XGB) model by applying the shapley additive explanations (SHAP) technique. The XGB algorithm presented its highest accuracy, with an average fivefold accuracy of 0.99% precision, 0.95% recall, and 0.97% F1 score. We evaluated the significance of each pulsatile feature's contribution to predictions, offering a more profound understanding of the model's behavior in distinguishing CSF lumen pixels with SHAP. Introducing a novel approach in the field, develop ML models offer comprehension into feature extraction and selection from PC-MRI pulsatile data. Moreover, the explained ML model offers novel and valuable insights to domain experts, contributing to an enhanced scholarly understanding of CSF dynamics.

Integrated finite element analysis and machine learning approach for propagation pressure prediction in hybrid Steel-CFRP subsea pipelines

Accurate prediction of the propagation pressure (PP) in hybrid steel-CFRP pipe systems presents a substantial challenge due to intricate interactions and complex collapse failure modes. An efficient FE-based algorithm is programmed using ANSYS to numerically estimate the PP of hybrid steel-CFRP pipe, subjected to external pressure. This study employs a machine learning (ML) framework, addressing the inherent complexity with a three-phase approach: Parameter Design, Buckle Propagation Analysis, and ML Model Development. The dataset, encompassing about two thousand observations with four key features, undergoes k-fold cross-validation and min-max normalization for robust ML performance. Five ML models—Random Forest (RF), K-Nearest Neighbors (KNN), Genetic Programming (GP), Multi-layer Perceptron (MLP), and Support Vector Machine (SVM)—are developed and evaluated. The results revealed a significant influence of Ds/ts, a three-phase relationship with ts/tc, and a substantial decrease in PPh/PPs with increasing σys/σuc, predominantly exhibiting U-shaped or dog-bone failure modes in different scenarios. Proven that GP, KNN, and RF are the superior performers, ranking ahead of SVM with Gaussian Kernel (SVM-GK), MLP, and SVM with Linear Kernel (SVM-LK). Statistical metrics, Taylor Diagram analysis, and comparisons with FE results emphasize the effectiveness of GP, KNN, and RF. Additionally, normality tests and feature importance analysis provide nuanced insights.

Comparison of Machine Learning Models to a Novel Score in the Identification of Patients at Low Risk for Diabetic Retinopathy

PurposeTo develop an easily applicable predictor of patients at low risk for diabetic retinopathy (DR). DesignAn experimental study on the development and validation of machine learning models and a novel retinopathy risk score to detect patients at low risk for diabetic retinopathy. SubjectsAll individuals aged 18 years or older who participated in the telemedicine retinal screening initiative through Temple University Health Systems from October 1, 2016 through December 31, 2020. The subjects must have documented evidence of their diabetes mellitus diagnosis as well as a documented hemoglobin A1C recorded in their chart within 6 months of the retinal screening photograph. MethodsThe charts of 1930 subjects (1590 evaluable) undergoing telemedicine screening for DR were reviewed, and 30 demographic and clinical parameters were collected. DR is a dichotomous variable where low risk is defined as no or mild retinopathy using the International Clinical Diabetic Retinopathy severity score. Five machine learning models (MLM) were trained to predict patients at low risk for DR using 1050 subjects and further underwent 10-fold cross validation to maximize its performance indicated by the area under the receiver operator characteristic curve (AUC). Additionally, a novel retinopathy risk score (RRS) is defined as the product of hemoglobin A1c closest to screening and years with diabetes mellitus. RRS was also applied to generate a predictive model. Main Outcome MeasuresThe performance of the trained machine learning models and the RRS model was compared using DeLong’s test. The models were further validated using a separate unseen test set of 540 subjects.The performance of the validation models were compared using DeLong’s test and chi-square tests. ResultsUsing the test set, the AUC for the RRS was not statistically different from four out of five MLM. The error rate for predicting low-risk patients using the RRS was significantly lower than the naive rate (0.097 vs 0.19; p<0.0001), and it was comparable to the error rates of the machine learning models. ConclusionThis novel retinopathy risk score is a potentially useful and easily deployable predictor of patients at low risk for DR.

Machine Learning Models for Predictive Maintenance in Industrial Engineering

Purpose: The general objective of this study was to investigate machine learning models for predictive maintenance in industrial engineering. Methodology: The study adopted a desktop research methodology. Desk research refers to secondary data or that which can be collected without fieldwork. Desk research is basically involved in collecting data from existing resources hence it is often considered a low cost technique as compared to field research, as the main cost is involved in executive’s time, telephone charges and directories. Thus, the study relied on already published studies, reports and statistics. This secondary data was easily accessed through the online journals and library.Findings: The findings reveal that there exists a contextual and methodological gap relating to machine learning models for predictive maintenance in industrial engineering. The research highlighted the transformative potential of machine learning models in optimizing predictive maintenance for industrial engineering, demonstrating significant reductions in unplanned downtime and maintenance costs. It identified the strengths of various machine learning approaches, such as supervised, unsupervised, and reinforcement learning, in predicting equipment failures and optimizing maintenance schedules. Despite the benefits, challenges such as data quality, integration complexity, and the need for specialized skills were noted. Future advancements in machine learning, IoT data, and computational power were expected to further enhance predictive maintenance systems, making them more accurate, efficient, and widely adopted across industries.Unique Contribution to Theory, Practice and Policy: The Systems Theory, Diffusion of Innovations Theory and Resource-Based View (RBV) Theory may be used to anchor future studies on machine learning models for predictive maintenance in industrial engineering. This study provided several recommendations that contributed to theory, practice, and policy. It emphasized the development of hybrid machine learning models, integration of domain-specific knowledge, and real-time data collection using IoT technologies. It suggested standardized data protocols and personnel training for better implementation and efficiency. Policy recommendations included regulatory frameworks, incentives for technology adoption, data sharing, and robust data privacy guidelines. These contributions aimed to enhance the accuracy and applicability of predictive maintenance models, improve industrial maintenance practices, and support technological innovation through supportive policies.

A Novel Machine-Learning Algorithm to Predict the Early Termination of Nutrition Support Team Follow-Up in Hospitalized Adults: A Retrospective Cohort Study

Background: For hospitalized adults, it is important to initiate the early reintroduction of oral food in accordance with nutrition support team guidelines. The aim of this study was to develop and validate a machine learning-based algorithm that predicts the early termination of medical nutritional therapy (the transition to oral feeding). Methods: This retrospective cohort study included consecutive adult patients admitted to the Hacettepe hospital (from 1 January 2018 to 31 December 2022). The outcome of the study was the prediction of an early transition to adequate oral feeding before discharge. The dataset was randomly (70/30) divided into training and test datasets. We used six ML algorithms with multiple features to construct prediction models. ML model performance was measured according to the accuracy, area under the receiver operating characteristic curve, and F1 score. We used the Boruta Method to determine the important features and interpret the selected features. Results: A total of 2298 adult inpatients who were followed by a nutrition support team for medical nutritional therapy were included. Patients received parenteral nutrition (1471/2298, 64.01%), enteral nutrition (717/2298, 31.2%), or supplemental parenteral nutrition (110/2298, 4.79%). The median (interquartile range) Nutritional Risk Screening (NRS-2002) score was 5 (1). Six prediction algorithms were used, and the artificial neural network and elastic net models achieved the greatest area under the ROC in all outcomes (AUC = 0.770). Ranked by z-value, the 10 most important features in predicting an early transition to oral feeding in the artificial neural network and elastic net algorithms were parenteral nutrition, surgical wards, surgical outcomes, enteral nutrition, age, supplemental parenteral nutrition, digestive system diseases, gastrointestinal complications, NRS-2002, and impaired consciousness. Conclusions: We developed machine learning models for the prediction of an early transition to oral feeding before discharge. Overall, there was no discernible superiority among the models. Nevertheless, the artificial neural network and elastic net methods provided the highest AUC values. Since the machine learning model is interpretable, it can enable clinicians to better comprehend the features underlying the outcomes. Our study could support personalized treatment and nutritional follow-up strategies in clinical decision making for the prediction of an early transition to oral feeding in hospitalized adult patients.

Developing machine learning models to predict the fly ash concrete compressive strength

Developing machine learning models to predict the fly ash concrete compressive strength

Understanding the synergistic geopolymerization mechanism of multiple solid wastes in ternary geopolymers

The use of ground granulated blast furnace slag (GBFS), fly ash (FA) and steel slag (SS) to form ternary geopolymers consumes several industrial solid wastes and produces high-performance construction materials. However, the unexplored synergistic geopolymerization mechanism of multiple solid wastes limits the design of high-performance ternary geopolymers. This study investigated the synergistic mechanism through testing the macroscopic performances such as unconfined compressive strength (UCS), final setting time (FST) and fluidity, and characterizing the microscopic structure of ternary geopolymers with different mixture proportions. In addition, the influence of variables on strength development of ternary geopolymers is analyzed by developing machine learning models. Results reveal that the highest strength of ternary geopolymers was achieved at a ratio of FA to GBFS to SS of 3:5:2, and alkali activator to precursor materials of 0.08. Addition of 20 % SS refines the pore network, reducing cumulative and detrimental pore volumes, thus enhancing compressive strength. In addition, the SS addition extends the heat generation peak during geopolymerization due to its slower reaction kinetics. Material components (Na/Al, Si/Al, Ca/Si) have a significant influence on UCS, with Na/Al ratio being the most sensitive variable. The study provides a new approach for utilization of bulk industrial wastes.

Concomitant Procedures, Black Race, Male Sex, and General Anesthesia Show Fair Predictive Value for Prolonged Rotator Cuff Repair Operative Time: Analysis of the NSQIP Database Using Machine Learning

PurposeThe purpose of this study is to develop machine learning models using the American College of Surgeons National Quality Improvement Program (ACS-NSQIP) database to predict prolonged operative time (POT) for rotator cuff repair (RCR). Furthermore, this study aims to use the trained machine learning (ML) models, cross-referenced with traditional multivariate logistic regression (MLR), to determine the key perioperative variables that may predict POT for RCR. MethodsData were obtained from a large, national database (NSQIP) from 2021. Patients with unilateral RCR procedures were included. Demographic, preoperative, and operative variables were analyzed. A multivariable logistic regression (MLR) model and various other machine learning techniques, including random forest (RF) and artificial neural network (ANN), were compared using area under the curve (AUC), calibration, Brier score, and decision curve analysis. Feature importance was identified from the overall best-performing model. ResultsA total of 6,690 patients met inclusion criteria. The random forest (RF) ML model had the highest AUC upon internal validation (0.706) and the lowest Brier score (0.15), outperforming the other models. The RF model also demonstrated strong performance upon assessment of the calibration curves (Slope = 0.86, Intercept = 0.08) and decision curve analysis. The model identified concomitant procedure, specifically labral repair and biceps tenodesis, as the most important variable for determining POT, followed by age <30 years, Black or African American race, male sex, and general anesthesia. ConclusionsDespite the advanced machine learning models used in this study, the NSQIP dataset was only able to fairly predict POT following RCR. The RF model identified concomitant procedures, specifically labral repair and biceps tenodesis, as the most important variables for determining POT. Additionally, demographic factors such as age <30 years, Black race, and general anesthesia were significant predictors. While male sex was identified as important in the RF model, the MLR model indicated that its predictive value is primarily in conjunction with specific procedures like biceps tenodesis and subacromial decompression.

Adaptive machine learning surrogate based multiobjective optimization for scavenging residual saltwater behind subsurface dams

High-fidelity physically based groundwater flow and solute transport models have been limited for seawater intrusion remediation design because of computationally intensive evolutionary algorithms. Data-driven machine learning approaches are promising to substitute expensive-cost groundwater numerical models within optimization due to computing efficiency. However, machine learning surrogates may accumulate error of forecasting and thereby result in infeasible optimal solutions. To achieve desired fidelity level, this study proposes a novel adaptive machine learning surrogate based multiobjective optimization method for coastal aquifer desalination. An adaptive modeling algorithm is newly introduced to iteratively retrain poorly-performed machine learning models and enhance predicting accuracy. The method is demonstrated to seek optimal extraction and injection strategies for scavenging residual saltwater trapped in an upstream aquifer behind a subsurface dam. Two conflicting objectives of minimizing the total extraction-injection rate and maximizing saltwater removal effectiveness are considered. Three machine learning models including artificial neural network (ANN), Gaussian process (GP) and response surface regression model (RSR) are developed to replace a high-fidelity seawater intrusion model for predicting chloride concentration and salinity mass. Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to derive Pareto fronts. Pareto optimal solutions obtained from machine learning models are compared against those from the seawater intrusion model. Results indicate that the developed machine learning models do not only have strong predicting capability, but also maintain good quality of Pareto optimal solutions, while achieving substantial computational saving up to 95%. Especially, the adaptively retrained RSR model inhibits error accumulation of forecasting and accelerates correct convergence to the true Pareto front. The proposed method is found superior performance in accurate convergence, widely-spread diversity as well as high computing efficiency than conventional methods.

Second opinion machine learning for fast-track pathway assignment in hip and knee replacement surgery: the use of patient-reported outcome measures

BackgroundThe frequency of hip and knee arthroplasty surgeries has been rising steadily in recent decades. This trend is attributed to an aging population, leading to increased demands on healthcare systems. Fast Track (FT) surgical protocols, perioperative procedures designed to expedite patient recovery and early mobilization, have demonstrated efficacy in reducing hospital stays, convalescence periods, and associated costs. However, the criteria for selecting patients for FT procedures have not fully capitalized on the available patient data, including patient-reported outcome measures (PROMs).MethodsOur study focused on developing machine learning (ML) models to support decision making in assigning patients to FT procedures, utilizing data from patients’ self-reported health status. These models are specifically designed to predict the potential health status improvement in patients initially selected for FT. Our approach focused on techniques inspired by the concept of controllable AI. This includes eXplainable AI (XAI), which aims to make the model’s recommendations comprehensible to clinicians, and cautious prediction, a method used to alert clinicians about potential control losses, thereby enhancing the models’ trustworthiness and reliability.ResultsOur models were trained and tested using a dataset comprising 899 records from individual patients admitted to the FT program at IRCCS Ospedale Galeazzi-Sant’Ambrogio. After training and selecting hyper-parameters, the models were assessed using a separate internal test set. The interpretable models demonstrated performance on par or even better than the most effective ‘black-box’ model (Random Forest). These models achieved sensitivity, specificity, and positive predictive value (PPV) exceeding 70%, with an area under the curve (AUC) greater than 80%. The cautious prediction models exhibited enhanced performance while maintaining satisfactory coverage (over 50%). Further, when externally validated on a separate cohort from the same hospital-comprising patients from a subsequent time period-the models showed no pragmatically notable decline in performance.ConclusionsOur results demonstrate the effectiveness of utilizing PROMs as basis to develop ML models for planning assignments to FT procedures. Notably, the application of controllable AI techniques, particularly those based on XAI and cautious prediction, emerges as a promising approach. These techniques provide reliable and interpretable support, essential for informed decision-making in clinical processes.

Plasma-based near-infrared spectroscopy for early diagnosis of lung cancer

Lung cancer (LC) continues to be a leading death cause in China, primarily due to late diagnosis. This study aimed to evaluate the effectiveness of using plasma-based near-infrared spectroscopy (NIRS) for LC early diagnosis. A total of 171 plasma samples were collected, including 73 healthy controls (HC), 73 LC, and 25 benign lung tumors (B). NIRS was utilized to measure the spectra of samples. Pre-processing methods, including centering and scaling, standard normal variate, multiplicative scatter correction, Savitzky-Golay smoothing, Savitzky-Golay first derivative, and baseline correction were applied. Subsequently, 4 machine learning (ML) algorithms, including partial least squares (PLS), support vector machines (SVM), gradient boosting machine, and random forest, were utilized to develop diagnostic models using train set data. Then, the predictive performance of each model was evaluated using test set samples. The study was conducted in 5 comparisons as follows: LC and HC, LC and B, B and HC, the diseased group (D) and HC, as well as LC, B and HC. Among the 5 comparisons, SVM consistently generated the best performance with a certain pre-processing method, achieving overall accuracy of 1.0 (kappa: 1.0) in the comparisons of LC and HC, B and HC, as well as D and HC. Pre-processing was identified as a crucial step in developing ML models. Interestingly, PLS demonstrated remarkable stability and relatively high predictive performance across the 5 comparisons, even though it did not achieve the top results like SVM. However, none of these algorithms were able to effectively distinguish B from LC. These findings indicate that the combination of plasma-based NIRS with ML algorithms is a rapid, non-invasive, effective, and economical method for LC early diagnosis.

Using the Height-for-Age Z-score (HAZ) to predict protein-energy malnutrition in patients with congenital epidermolysis bullosa through the application of machine learning methods

Congenital epidermolysis bullosa (СЕВ) is a group of genetically and clinically heterogeneous diseases characterized by a tendency to form blisters and/or erosions on the skin and mucous membranes with minimal trauma. Nutritional deficiency stands as the most common complication observed in EB, with its development influenced by a multitude of contributing factors.Purpose of the study. To construct a predictive model for the development of protein-energy malnutrition in patients with various forms of EB, identify the main features affecting the sensitivity of the predictive model, and evaluate the model’s validity based on retrospective data on the presence of protein-energy malnutrition in this patient category.Methods. The study involved 101 patients aged 3 to 18 years with simplex (n=25), junctional (n=10), and dystrophic (n=66) СEB. The Birmingham EB Severity Score, laboratory and anthropometric parameters, as well as data on the presence of gastrointestinal complications, were used for the analysis of disease progression and predictive model construction. The Scikit-learn library of the programming language Python was utilized for building the machine learning model.Results. In the construction of the predictive model, the RandomForestClassifier model showed the best results. The developed machine learning model can correctly determine whether a patient has chronic protein-energy malnutrition (class 1, HAZ &lt; –2) or not (class 0, HAZ &gt; –2) with an accuracy of 92%, sensitivity of 85.7%, and specificity of 100%.Conclusions. The machine learning model presented in this study predicts the values of the Height-for-Age Z-score (HAZ) and can have practical significance in medical practice and clinical research. The model can be used for early diagnosis of protein-energy malnutrition in patients with EB, which may allow healthcare professionals to timely start nutritional support and prevent possible complications of the disease, as well as develop individual nutrition and treatment plans for patients.

Development and optimization of machine learning models for estimation of mechanical properties of linear low-density polyethylene

A hybrid methodology was developed and implemented for estimation of polymeric mechanical properties in rotational moulding process. The considered polymer in this study is linear low-density polyethylene, known as LLDPE, which has extensive application in plastic industry. The mechanical properties of the polymer were assessed and correlated to the oven residence time to build the predictive model of moulding process. A tiny dataset containing only 25 data rows via a number of machine learning models were assessed. Oven residence time is the only input, while the LLDPE's properties including tensile strength, impact strength, and flexure strength are the outputs considered in the machine learning models. We used tree-based ensemble methods for modeling in this work and they are tuned using FA (Firefly Algorithm) optimizer to find optimal hyper-parameters of them. Finally, the optimal models had shown a great performance to predict the output accurately. For tensile strength, the best model (FA-ET) has an R2 value of 0.9994, this score is 0.9995 for impact strength and 0.9968 for flexure strength. The tree-based models tuned in this study revealed to be robust in estimation of polymeric properties and can be used to obtain the products with the best quality.

Deep learning from three-dimensional Lithium-ion battery multiphysics model Part II: Convolutional neural network and long short-term memory integration

Lithium-ion (Li-ion) batteries have emerged as a cornerstone of electric vehicles (EVs), enabling the road transportation towards net zero. The success of electric vehicles largely hinges on the battery performance and safety. It is challenging to test and predict battery performance and safety issues by conventional methods, which are usually time-consuming and expensive, involving significant human and measurement errors. To enable the quick estimation of battery performance and safety, we developed three data-driven machine learning (ML) models, namely a convolutional neural network (CNN), a long short-term memory (LSTM), and a CNN-LSTM to predict battery discharge curves and local maximum temperature (hot spot) under various operating conditions. The developed ML models mitigated data scarcity by employing a three-dimensional multi-physics Li-ion battery model to generate enormous and diverse high-quality data. It was found the CNN-LSTM model outperforms the others and achieved high accuracy of 98.68% to learn discharge curves and battery maximum temperature, owing to the integration of spatial and sequential feature extraction. The battery safety can be improved by comparing the predicted maximum battery temperature against safe temperature threshold. The proposed data development and data-driven ML models are of great potential to provide digital tools for engineering high-performance and safe EVs.