Machine Learning Models For Prediction Research Articles

Performance comparison of machine learning and deep learning models for supply chain tier order quantity prediction: Emphasis on tree-based and CNN-BILSTM approaches

Performance comparison of machine learning and deep learning models for supply chain tier order quantity prediction: Emphasis on tree-based and CNN-BILSTM approaches

Development of Time-Aggregated Machine Learning Model for Relapse Prediction in Pediatric Crohn's Disease.

Pediatric Crohn's disease (CD) easily progresses to an active disease compared to adult CD, making it important to predict and minimize CD relapses. However, prediction of relapse at various time points (TPs) during pediatric CD remains understudied. We aimed to develop a real-time aggregated model to predict pediatric CD relapse in different TPs and time windows (TWs). This retrospective study was conducted on children diagnosed with CD between 2015 and 2022 at Severance Hospital. Laboratory test results and demographic data were collected starting at 3 months after diagnosis, and cohorts were formed using data from six different TPs at 1-month intervals. Relapse-defined as a pediatric CD activity index ≥30 points-was predicted, and TWs were 3-7 months with 1-month intervals. The feature importance of the variables in each setting was determined. Data from 180 patients were used to construct cohorts corresponding to the TPs. We identified the optimal TP and TW to reliably predict pediatric CD relapse with an area under the receiver operating characteristic curve score of 0.89 when predicting with a 3-month TW at a 3-month TP. Variables such as C-reactive protein levels and lymphocyte fraction were found to be important factors. We developed a time-aggregated model to predict pediatric CD relapse in multiple TPs and TWs. This model identified important variables that predicted relapse in pediatric CD to support real-time clinical decision making.

Development of a simplified model and nomogram for the prediction of pulmonary hemorrhage in respiratory distress syndrome in extremely preterm infants

BackgroundPulmonary hemorrhage (PH) in respiratory distress syndrome (RDS) in extremely preterm infants exhibits a high mortality rate and poor long-term outcomes. The aim of the present study was to develop a machine learning (ML) predictive model for RDS with PH in extremely preterm infants.MethodsWe performed a retrospective analysis of extremely preterm infants with RDS at the Children’s Hospital of Soochow University between January 2015 and January 2021. We applied three ML algorithms—logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost)—to evaluate the performance of each model using the area under the curve (AUC), and developed a predictive model based on the optimal model. We calculated SHapley Additive exPlanations (SHAP) values to determine variables importance and show visualization results, and constructed a nomogram for individualized prediction.ResultsA total of 309 patients with RDS were enrolled, including 48 (15.5%) with PH. A total of 29 variables were collected, including demographic and clinical characteristics, laboratory data, and image classification. According to the AUC values, the RF model performed best (AUC = 0.868). Based on the SHAP values, the top five important variables in the RF model were gestational age, PaO2/FiO2, birth weight, mean platelet volume, and Apgar score at 5 min.ConclusionsOur study showed that the RF model could be used to predict the risk of PH in RDS in extremely preterm infants. The nomogram provides clinicians with an effective tool for early warning and timely management.

A comprehensive framework of the decomposition-based hybrid method for ultra-short-term wind power forecasting with on-site application

Ultra-short-term wind power forecasting (UWPF) is crucial for the large-scale grid connection of wind energy. The state grid in China has strict multi-step forecasting requirements, which pose challenges to on-site efficiency and accuracy. This paper proposes a comprehensive framework of the hybrid method for UWPF with on-site application, consisting of data decomposition, model prediction, and post-processing. In the first stage, rolling decomposition and feature reconstruction are employed to decompose the wind power data into sub-components without future information leakage. The feature extraction and model matching processes are then performed to make full use of different machine learning prediction models and distinct characteristics of wind power components. Finally, a novel error tracking strategy is proposed to enable real-time error correction for multi-step forecasting by capturing the fluctuation characteristics of wind power data. The proposed framework is evaluated on two field wind power datasets through comparative experiments with five benchmark methods and nine reference methods. The experimental results demonstrate that (a) Each module of the proposed method effectively contributes to improving forecasting accuracy. (b) The proposed method significantly outperforms traditional machine learning methods in both single-step and multi-step forecasts, indicating its capability to handle practical UWPF tasks effectively.

Evaluation of machine learning and deep learning models for daily air quality index prediction in Delhi city, India.

The air quality index (AQI), based on criteria for air contaminants, is defined to provide a shared vision of air quality. As air pollution continues to rise in global cities due to urbanization and climate change, air pollution monitoring and forecasting models for effective air quality monitoring that gather and forecast information about air pollution concentration are essential in every city. Air quality predictions have evolved to be more helpful for management. Recently, better performance and ability have developed due to the involvement of machine learning (ML) and artificial intelligence (AI) in forecasting air quality in urban cities in India. This paper focuses on air pollution as a significant ecological problem that directly impacts human health and the distribution of an environmental system in urban areas. Hence, we have developed advanced models for daily AQI forecasting to understand the air effluence level in the upcoming days. In this research, six data-driven models have been developed and implemented for daily AQI forecasting in the study area; it is crucial for understanding the future air pollution levels to plan and control air pollution in the entire city. The developed model is applied to air quality datasets. A comparison of the performance of ML models tested here indicates that the XGBoost algorithm achieves the highest coefficient of determination (R2) and root-mean-square deviation (RMSE) value of 0.99 and lower values value of 4.65 than other models in the testing phase. The results of the artificial neural network (ANN) algorithm are slightly lower than the extreme gradient boosting (XGBoost model); the ANN model results are as R2, mean squared error (MSE), and RMSE values of 0.99, 13.99, and 198.88, respectively. All the models were subjected to a ten-fold cross-validation model. However, the RF cross-validation model outperforms other models; the RF model result shows the R2, RMSE, and MSE values of 0.99, 3.64, and 4.12, respectively. This study also employed two interpretable models, namely feature importance analysis and Shapley additive explanation (SHAP), to evaluate both the global and local methods in a manner that is independent of specific ML models. The feature importance shows that particle matter (PM) 2.5, PM10, carbon monoxide (CO), and nitrogen oxides (NOx) were the most influential variables. The results determined that such novel DL and ML models may improve the accuracy of AQI forecasts and understanding of air pollution, particularly in metropolitan cities.

A machine learning prediction model for Cardiac Amyloidosis using routine blood tests in patients with left ventricular hypertrophy

Current approaches for cardiac amyloidosis (CA) identification are time-consuming, labor-intensive, and present challenges in sensitivity and accuracy, leading to limited treatment efficacy and poor prognosis for patients. In this retrospective study, we aimed to leverage machine learning (ML) to create a diagnostic model for CA using data from routine blood tests. Our dataset included 6,563 patients with left ventricular hypertrophy, 261 of whom had been diagnosed with CA. We divided the dataset into training and testing cohorts, applying ML algorithms such as logistic regression, random forest, and XGBoost for automated learning and prediction. Our model’s diagnostic accuracy was then evaluated against CA biomarkers, specifically serum-free light chains (FLCs). The model’s interpretability was elucidated by visualizing the feature importance through the gain map. XGBoost outperformed both random forest and logistic regression in internal validation on the testing cohort, achieving an area under the curve (AUC) of 0.95 (95%CI: 0.92–0.97), sensitivity of 0.92 (95%CI: 0.86–0.98), specificity of 0.95 (95%CI: 0.94–0.97), and an F1 score of 0.89 (95%CI: 0.85–0.92). Its performance was also superior to the serum FLC-kappa and FLC-lambda combination (AUC of 0.88). Furthermore, XGBoost identified unique biomarker signatures indicative of multisystem dysfunction in CA patients, with significant changes in eGFR, FT3, cTnI, ANC, and NT-proBNP. This study develops a highly sensitive and accurate ML model for CA detection using routine clinical laboratory data, effectively streamlining diagnostic procedures, and providing valuable clinical insights and guiding future research into disease mechanisms.

Machine learning-based prediction of elevated N terminal pro brain natriuretic peptide among US general population.

Natriuretic peptide-based pre-heart failure screening has been proposed in recent guidelines. However, an effective strategy to identify screening targets from the general population, more than half of which are at risk for heart failure or pre-heart failure, has not been well established. This study evaluated the performance of machine learning prediction models for predicting elevated N terminal pro brain natriuretic peptide (NT-proBNP) levels in the US general population. Individuals aged 20-79years without cardiovascular disease from the nationally representative National Health and Nutrition Examination Survey 1999-2004 were included. Six prediction models (two conventional regression models and four machine learning models) were trained with the 1999-2002 cohort to predict elevated NT-proBNP levels (>125pg/mL) using demographic, lifestyle, and commonly measured biochemical data. The model performance was tested using the 2003-2004 cohort. Of the 10237 individuals, 1510 (14.8%) had NT-proBNP levels >125pg/mL. The highest area under the receiver operating characteristic curve (AUC) was observed in SuperLearner (AUC [95% CI]=0.862 [0.847-0.878], P<0.001 compared with the logistic regression model). The logistic regression model with splines showed a comparable performance (AUC [95% CI]=0.857 [0.841-0.874], P=0.08). Age, albumin level, haemoglobin level, sex, estimated glomerular filtration rate, and systolic blood pressure were the most important predictors. We found a similar prediction performance even after excluding socio-economic information (marital status, family income, and education status) from the prediction models. When we used different thresholds for elevated NT-proBNP, the AUC (95% CI) in the SuperLearner models 0.846 (0.830-0.861) for NT-proBNP>100pg/mL and 0.866 (0.849-0.884) for NT-proBNP>150pg/mL. Using nationally representative data from the United States, both logistic regression and machine learning models well predicted elevated NT-proBNP. The predictive performance remained consistent even when the models incorporated only commonly available variables in daily clinical practice. Prediction models using regularly measured information would serve as a potentially useful tools for clinicians to effectively identify targets of natriuretic-peptide screening.

Optimizing flexural strength of RC beams with recycled aggregates and CFRP using machine learning models.

This paper investigates the flexural bearing behavior of reinforced concrete beams through experimental analysis and advanced machine learning predictive models. The primary problem centers around understanding how varying compositions of construction materials, particularly the inclusion of recycled aggregates and carbon fiber-reinforced polymer (CFRP), affect the structural performance of concrete beams. Eight beams, including those with natural aggregates, recycled aggregates, fly ash, and CFRP, were tested. The study employs state-of-the-art machine learning frameworks, including Random Forest Regressor (RFR), XGBoost (XGB), and LightGBM (LGBM). The formation of these models involved data acquisition from experiments, preprocessing of key input features (such as rebars area, cement portion, recycled and natural aggregate masses, silica fume, fly ash, compressive strength, and CFRP presence), model selection, and hyperparameter tuning using Pareto optimization. The models were then evaluated using performance metrics like Mean Squared Error (MSE), Mean Absolute Error (MAE), and coefficient of determination (R2). Outputs focus on load-induced deflection and mid-span displacement. With a dataset of 4851 samples, the optimized models demonstrated excellent performance. The experimental results revealed substantial enhancements in both compressive strength and load-bearing capacity, notably observed in beams incorporating 70% recycled aggregate and 10% silica fume. These beams exhibited a remarkable increase in compressive strength of up to 53.03% and a 7% boost in load-bearing capacity compared to those without recycled aggregate. By integrating experimental analysis with advanced computational techniques, this study advances the understanding of eco-friendly construction materials and their performance, shedding light on the intricate interactions between sustainable construction materials and the flexural bearing behavior of beams.

Application of Machine Learning and Hydrological Models for Drought Evaluation in Ungauged Basins Using Satellite-Derived Precipitation Data

Drought is a complex environmental hazard to ecosystems and society. Decision-making on drought management options requires evaluating and predicting the extremity of future drought events. In this regard, quantifiable indices such as the standardized precipitation index (SPI), the standardized precipitation evapotranspiration index (SPEI), and the standardized streamflow index (SSI) have been commonly used to characterize meteorological and hydrological drought. In general, the estimation and prediction of the indices require an extensive range of precipitation (SPI and SPEI) and discharge (SSI) datasets in space and time domains. However, there is a challenge for long-term and spatially extensive data availability, leading to the insufficiency of data in estimating drought indices. In this regard, this study uses satellite precipitation data to estimate and predict the drought indices. SPI values were calculated from the precipitation data obtained from the Centre for Hydrometeorology and Remote Sensing (CHRS) data portal for a study water basin. This study employs a hydrological model for calculating discharge and drought in the overall basin. It uses random forest (RF) and support vector regression (SVR) as machine learning models for SSI prediction for time scales of 1- and 3-month periods, which are widely used for establishing interactions between predictors and predictands that are both linear and non-linear. This study aims to evaluate drought severity variation in the overall basin using the hydrological model and compare this result with the machine learning model’s results. The results from the prediction model, hydrological model, and the station data show better correlation. The coefficients of determination obtained for 1-month SSI are 0.842 and 0.696, and those for the 3-month SSI are 0.919 and 0.862 in the RF and SVR models, respectively. These results also revealed more precise predictions of machine learning models in the longer duration as compared to the shorter one, with the better prediction result being from the SVR model. The hydrological model-evaluated SSI has 0.885 and 0.826 coefficients of determination for the 1- and 3-month time durations, respectively. The results and discussion in this research will aid planners and decision-makers in managing hydrological droughts in basins.

Machine learning predictive model for aspiration risk in early enteral nutrition patients with severe acute pancreatitis

BackgroundThe aim of this study was to build and validate a risk prediction model for aspiration in severe acute pancreatitis patients receiving early enteral nutrition (EN) by identifying risk factors for aspiration in these patients. MethodsThe risk factors for aspiration were analyzed to build a prediction model based on the data collected from 339 patients receiving enteral nutrition. Subsequently, we used six machine learning algorithms and the model was validated by the area under the curve. ResultsIn this study, the collected data were divided into two groups: a training cohort and a validation cohort. The results showed that 28.31 % (77) of patients had aspiration and 71.69 % (195) of patients had non-aspiration in training cohort. Moreover, age, consciousness, mechanical ventilation, aspiration history, nutritional risk and number of comorbidities were included as predictive factors for aspiration in patients receiving EN. The XGBoost model is the best of all machine learning models, with an AUROC of 0.992 and an F1 value of 0.902. The specificity and accuracy of XGBoost are higher than those of traditional logistic regression. ConclusionIn accordance with the predictive factors, XGBoost model, characterized by excellent discrimination and high accuracy, can be used to clinically identify severe acute pancreatitis patients with a high risk of enteral nutrition aspiration. Relevance to clinical practiceThis study contributed to the development of a predictive model for early enteral nutrition aspiration in severe acute pancreatitis patients during hospitalization that can be shared with medical staff and patients in the future. No patient or public contributionThis is a retrospective cohort study, and no patient or public contribution was required to design or undertake this research.

Research on National Short-Term Precipitation Forecast Method Based on Machine Learning

Abstract. Climate change is closely related to human lives. With the development of information technology, meteorological data gradually show the characteristics of big data. The purpose of this study is to compare the application effects of different machine learning models in short-term precipitation prediction, to improve the accuracy and efficiency of prediction. Based on the annual average precipitation data of prefecture-level cities in China from 1990 to 2022, this paper uses Random Forest, eXtreme Gradient Boosting and Neural Networks to construct prediction models, and comprehensively evaluate them. In this study, the data is preprocessed to ensure the data quality of the input model. Next, the data is fed into three machine learning algorithm models, Random Forest, eXtreme Gradient Boosting and Neural Networks. Finally, the prediction performance of each model is evaluated by various indexes. In this paper, it is found that the Random Forest model has the best performance, and its prediction accuracy is higher than the other two models, which has great application potential in the field of short-term precipitation prediction. This study shows that a reasonable selection of machine learning methods and optimization of model parameters can effectively improve the accuracy of short-term precipitation prediction. This paper provides some empirical evidence for precipitation prediction, which will help to make more effective decisions in dealing with extreme weather events and climate change challenges in the future.

Research on National Short-Term Precipitation Forecast Method Based on Machine Learning

Abstract. Climate change is closely related to human lives. With the development of information technology, meteorological data gradually show the characteristics of big data. The purpose of this study is to compare the application effects of different machine learning models in short-term precipitation prediction, to improve the accuracy and efficiency of prediction. Based on the annual average precipitation data of prefecture-level cities in China from 1990 to 2022, this paper uses Random Forest, eXtreme Gradient Boosting and Neural Networks to construct prediction models, and comprehensively evaluate them. In this study, the data is preprocessed to ensure the data quality of the input model. Next, the data is fed into three machine learning algorithm models, Random Forest, eXtreme Gradient Boosting and Neural Networks. Finally, the prediction performance of each model is evaluated by various indexes. In this paper, it is found that the Random Forest model has the best performance, and its prediction accuracy is higher than the other two models, which has great application potential in the field of short-term precipitation prediction. This study shows that a reasonable selection of machine learning methods and optimization of model parameters can effectively improve the accuracy of short-term precipitation prediction. This paper provides some empirical evidence for precipitation prediction, which will help to make more effective decisions in dealing with extreme weather events and climate change challenges in the future.

Predicting Fear of Breast Cancer Recurrence in women five years after diagnosis using Machine Learning and healthcare reimbursement data from the French nationwide VICAN survey

ObjectiveA major concern for cancer survivors after treatment is the Fear of Cancer Recurrence (FCR), which is the fear that cancer will reappear or progress. This fear can be exacerbated by medical uncertainty about the future, leading to harmful obsession and having a negative impact on quality of life. This study aims to develop a predictive Machine Learning (ML) model using healthcare reimbursement data to better predict FCR and understand the factors influencing FCR in women with breast cancer five years after their diagnosis. Materials and MethodsWe used data from the VICAN (VIe après le CANcer) survey to propose an interpretable model to identify patients at risk of developing clinical FCR. The reimbursement data for each patient were analyzed beyond the first two years of treatment, excluding the initial phase influenced by the cancer curative therapeutic process. Data experiments were conducted, including the use of algorithms such as Random Forest, Support Vector Machines, Gradient Boosting, eXtreme Gradient Boosting, and Multilayer Perceptron. The AUC was used to choose the optimal model. ResultsThe dataset is composed of 918 patients classified regarding FCR. The experimental design incorporated classification levels of medications, biological and medical procedures. Subsequently, data was generated for two experiments, facilitating examination at the ultimate healthcare classification level in Experiment 1, while Experiment 2 targeted the penultimate classification level. Overall, the best-performing model achieved an AUC of 66%. This demonstrates some effectiveness of the algorithms in detecting patients at risk of developing clinical FCR and encourages further investigations to enhance the model's performance and assess its generalizability. ConclusionML applied to reimbursement data has shown promise in predicting FCR, aiding in the identification of patients at risk of developing it. The results pave the way for personalized prevention and intervention strategies, representing a significant advancement in postcancer care focusing on the needs of survivors.

Abstract B075: Detection of early gastric carcinoma by micronuclei DNA from erythrocytes

Abstract Background: Advancements in non-invasive diagnostic technologies for gastric carcinoma (GC), including serum microRNA and cell-free DNA assays, have shown promise but still face challenges in achieving consistent early detection. Micronuclei (MN), extranuclear bodies harboring fragmented chromosomal segments, are indicative of genomic instability and have been employed in genotoxicity assessments and cancer diagnostics. We have developed a novel technique (WO2021/228246 A1) for the purification and characterization of micronuclei DNA (MN-DNA) extracted from erythrocytes in peripheral blood. Here, we evaluated the potential of micronuclei DNA (MN-DNA) from 2ml peripheral blood for early gastric carcinoma detection and explored its application for early cancer diagnosis. Methods: We assembled a clinical cohort consisting of 425 healthy donors (HDs) and 282 gastric cancer patients, from whom 1-2 mL peripheral blood samples were collected. Participants were randomly allocated into training, validation, and independent test cohorts in a 7:2:1 ratio, ensuring similar distributions of cancer stages and gender. MN-DNA was isolated from erythrocytes and subjected to comprehensive whole-genome sequencing. Predictive machine learning models were constructed utilizing distinct tumor-associated MN-DNA features identified through genome-wide analysis to distinguish between HDs and GC. Results: Comprehensive genome-wide analysis of MN-DNA demonstrated pronounced disparities in distribution patterns between HDs and GC patients. The predictive model built on MN-DNA features achieved an AUC of 97.69% (95%CI: 95.57- 99.81%), with an 96.47% specificity and 91.07% sensitivity in the independent test cohort. The model identified early GC (stage 0-I) and advanced GC (stage II-IV) with sensitivities of 95.24% or 78.57%, respectively. Conclusions: Our results demonstrate that MN-DNA from 1-2mL peripheral blood enable accurate detection of GC, especially in early GC. As a type of cytoplasmic DNA, MN-DNA can provide a valuable tool for early cancer detection, offering different underlying mechanisms compared to current methods. Research sponsor: Timing Biotech. Citation Format: Yuehua Han, Xingyun Yao, Haobo Sun, Cheng Fang, Peiwei Li, Honghao Liang, Jie Jin, Xiaofei Gao. Detection of early gastric carcinoma by micronuclei DNA from erythrocytes [abstract]. In: Proceedings of the AACR Special Conference: Liquid Biopsy: From Discovery to Clinical Implementation; 2024 Nov 13-16; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(21_Suppl):Abstract nr B075.

QTc interval prolongation impact on in-hospital mortality in acute coronary syndromes patients using artificial intelligence and machine learning

BackgroundPrediction of mortality in hospitalized patients is a crucial and important problem. Several severity scoring systems over the past few decades and machine learning models for mortality prediction have been developed to predict in-hospital mortality. Our aim in this study was to apply machine learning (ML) algorithms using QTc interval to predict in-hospital mortality in ACS patients and compare them to the validated conventional risk scores.ResultsThis study was retrospective, using supervised learning, and data mining. Out of a cohort of 500 patients admitted to a tertiary care hospital from September 2018 to August 2020, who presented with ACS. Prediction models for in-hospital mortality in ACS patients were developed using 3 ML algorithms. We employed the ensemble learning random forest (RF) model, the Naive Bayes (NB) model and the rule-based projective adaptive resonance theory (PART) model. These models were compared to one another and to two conventional validated risk scores; the Global Registry of Acute Coronary Events (GRACE) risk score and Thrombolysis in Myocardial Infarction (TIMI) risk score. Out of the 500 patients included in our study, 164 (32.8%) patients presented with unstable angina, 148 (29.6%) patients with non-ST-elevation myocardial infarction (NSTEMI) and 188 (37.6%) patients were having ST-elevation myocardial infarction (STEMI). 64 (12.8%) patients died in-hospital and the rest survived. Performance of prediction models was measured in an area under the receiver operating characteristic curve (AUC) ranged from 0.83 to 0.93 using all available variables compared to the GRACE score (0.9 SD 0.05) and the TIMI score (0.75 SD 0.02). Using QTc as a stand-alone variable yielded (0.67 SD 0.02) with a cutoff value 450 using Bazett’s formula, whereas using QTc in addition to other variables of personal and clinical data and other ECG variables, the result was 0.8 SD 0.04. Results of RF and NB models were almost the same, but PART model yielded the least results. There was no significant difference of AUC values after replacing the missing values and applying class balancer.ConclusionsThe proposed method can effectively predict patients at high risk of in-hospital mortality early in the setting of ACS using only clinical and ECG data. Prolonged QTc interval can be used as a risk predictor of in-hospital mortality in ACS patients.

A comparison of statistical and machine learning models for spatio-temporal prediction of ambient air pollutant concentrations in Scotland

The spatio-temporal prediction of air pollutant concentrations is vital for assessing regulatory compliance and for producing exposure estimates in epidemiological studies. Numerous approaches have been utilised for making such predictions, including land use regression models, additive models, spatio-temporal smoothing models and machine learning prediction algorithms. However, relatively few studies have compared the predictive performance of these models thoroughly, which is one of the novel contributions of this paper. For the specific challenge of predicting monthly average concentrations of NO2, PM10 and PM2.5 in Scotland, we find that random forests typically outperform (or are as good as) more traditional statistical prediction approaches. Additionally, we utilise the best performing model to provide a new data resource, namely, predictions of monthly average concentrations (with uncertainty quantification) of the above pollutants on a regular 1 km2 grid for all of Scotland between 2016 and 2020.

Superior forecasting with simple AR(1) models in a low-volatility environment: evidence from the CAT bond market

AbstractIn the recent literature on asset pricing, advanced machine learning methods often show better predictive quality than simple linear regression models. In this context, machine learning prediction models for bond premiums are usually based on those predictors that have proved to be particularly relevant in explanatory models. However, these models do not take into account that historical premiums of assets with particularly low premium volatilities already contain a high degree of information about future premiums. With this in mind, we consider catastrophe bonds, whose secondary market premiums exhibit low volatility, and include historical premiums in our forecasting models in addition to the usual predictors. In this way, the predictive accuracy of the linear regression is significantly increased and is comparable to that of advanced machine learning methods. Remarkably, a simple linear AR(1) model without additional predictors achieves the highest predictive performance.

Prediction Model for the Chloride Ion Permeability Resistance of Recycled Aggregate Concrete Based on Machine Learning

The chloride ion permeability resistance of recycled aggregate concrete (RAC) is influenced by multiple factors, and the prediction model for this resistance based on machine learning is still limited. In the paper, six impact factors (IFs), including the carbonation of recycled coarse aggregates (YN), the replacement ratio of recycled coarse aggregates (r), the bending load level (L), the carbonation time (t) and temperature (T) of RAC, and the replacement ratio of carbonated recycled fine aggregates (f), were considered to conduct a chloride penetration test on RAC. Based on the experimental data, four algorithms, including artificial neural network (ANN), support vector machine (SVM), random forest (RF) and extreme gradient boosting (XGBoost), were adopted to establish the machine learning prediction models and study the relationships between the electric flux of RAC and the IFs. The results showed that the predicted values of all four models were in good agreement with the experimental values, and the XGBoost model had the best prediction performance on the testing set. Based on the XGBoost model, the LIME method was adopted to solve the interpretability problem in the prediction process. The importance ranking of IFs on the electric flux was r &gt; t &gt; f &gt; T &gt; L &gt; YN. A graphical user interface (GUI) was developed based on Python 3.8 software to facilitate the use of machine learning models for the chloride ion permeability resistance of RAC. The research results can provide an accurate prediction of the electric flux of RAC.

Random forest models reveal academic and financial factors outweigh demographics in predicting completion of a year-round veterinary program.

The purpose of this study was to develop random forest classifier models (a type of supervised machine learning algorithm) that could (1) predict students who will or will not complete the DVM degree requirements and (2) identify the top predictors for academic success and completion of the DVM degree. The study utilized Ross University School of Veterinary Medicine student records from 2013 to 2022. Twenty-four variables encompassing demographic (eg, age, race), academic (eg, grade point average), and financial aid (eg, outstanding balances) data were assessed in 11 cross-validated random forest machine learning models. One model was built assessing all years of data and 10 individual models were developed for each enrollment year to compare how the top predictors of success varied among the years. Consistently, only academic and financial factors were identified as being features of importance (predictors) in all models. Demographic factors such as race were not important for predicting student success. All models performed very well to excellently based on multiple performance metrics including accuracy, ranging from 96.1% to 99%, and the areas under the receiver operating characteristic curves, ranging from 98.1% to 99.9%. The random forest algorithm is a powerful machine learning prediction model that performs well with veterinary student academic records and is customizable such that variables important to each veterinary school's student population can be assessed. Identifying predictors of success as well as at-risk students is essential for providing targeted curricular interventions to increase retention and achieve timely completion of a DVM degree.

Abstract 4122612: Validation of a Machine Learning Model for Fetal Echocardiographic Prediction of Critical Coarctation of the Aorta

Background: Current fetal echocardiographic (F-echo) metrics have inadequate specificity for confident prediction of neonatal critical coarctation of the aorta (CoA). Using single-center data, our machine learning model for the prediction of fetal CoA demonstrated improved accuracy compared to published F-echo metrics for critical CoA assessment. External validation of this model is needed. Aim: Validate a machine learning F-echo predictive model for CoA with an external patient cohort. Methods: Initial model training and testing were performed using retrospective single center data on 9 F-echo measurements for patients with prenatal concern for CoA. A random forest classifier with 80:20 split and 5-fold cross-validation predicted CoA intervention within 30 days of life. A SHapley Additive exPlanations (SHAP) analysis assessed the marginal contribution of each feature. The model was retrained using the 5 most influential F-echo features. External validation for this model was then performed using patients with prenatal concern for CoA retrospectively collected at a partner institution. Results: Inclusion criteria were met by 132 patients in the initial cohort and 64 patients in the external validation cohort, of whom 44% (n=58) and 25% (n=16) respectively had CoA requiring intervention. SHAP analysis for both cohorts demonstrated transverse to descending aorta angle as the most influential feature, followed by ascending to descending aorta angle (Figure 1A). Using internal cross-validation on the initial cohort, the area under the receiver operating characteristic curve (AUC) was 0.93 ± 0.05 (sensitivity 0.97, specificity 1.0) with an F1 score of 0.97 ± 0.03. Validation of the model with the external cohort produced an AUC of 0.87 (sensitivity 0.81, specificity 1.0) and an F1 of 0.90 (Figure 1B). Conclusions: A random forest classifier using F-echo features predicted neonatal critical CoA with higher accuracy than previously published metrics. The model maintained high accuracy when validated with an external patient cohort. Arch angles most significantly impacted the model’s accuracy. Future directions include prospective validation and converting the model to a distributable clinical calculator.