A Novel Active Learning Technique for Fetal Health Classification Based on Xgboost Classifier DOI
Kaushal Bhardwaj,

Niyati Goyal,

Bhavika Mittal

et al.

Published: Jan. 1, 2024

Ensuring safe pregnancy and reducing maternal infant mortality rates require addressing factors that affect fetal health. The application of machine learning algorithms in monitoring health helps to improve the chances timely intervention better outcomes case any possible issues fetuses. Existing studies offered aid this issue typically train models using a significant portion dataset, ranging mostly around 75%-80%. In work, we propose novel solution implementing an active technique identify most informative data samples for training model leading high accuracy with limited number samples. It employs query function built upon uncertainty diversity criteria which are derived based on properties XGBoost classifier. For deriving soft probabilities obtained unlabelled used while distance among feature space is utilized criteria. proposed approach shows superior performance comparison all state-of-the-art methods. Through analysis experimentation, achieves average higher than 99% by utilizing less 20% dataset training. This demonstrates its efficacy potential monitoring.

Language: Английский

An explainable stacking-based approach for accelerating the prediction of antidiabetic peptides DOI
Farwa Arshad, Saeed Ahmed, Aqsa Amjad

et al.

Analytical Biochemistry, Journal Year: 2024, Volume and Issue: 691, P. 115546 - 115546

Published: April 25, 2024

Language: Английский

Citations

1
Construction and SHAP interpretability analysis of a risk prediction model for feeding intolerance in preterm newborns based on machine learning DOI Creative Commons
Hui Xu,

Xingwang Peng,

Ziyu Peng

et al.

BMC Medical Informatics and Decision Making, Journal Year: 2024, Volume and Issue: 24(1)

Published: Nov. 18, 2024

To construct a highly accurate and interpretable feeding intolerance (FI) risk prediction model for preterm newborns based on machine learning (ML) to assist medical staff in clinical diagnosis. In this study, sample of 350 hospitalized were retrospectively analysed. First, dual feature selection was conducted identify important variables construction. Second, ML models constructed the logistic regression (LR), decision tree (DT), support vector (SVM) eXtreme Gradient Boosting (XGBoost) algorithms, after which random sampling tenfold cross-validation separately used evaluate compare these optimal model. Finally, we apply SHapley Additive exPlanation (SHAP) framework analyse decision-making principles expound upon factors affecting FI their modes action. The accuracy XGBoost 87.62%, area under curve (AUC) 92.2%. After application cross-validation, 83.43%, AUC 89.45%, significantly better than those other models. Analysis with SHAP showed that history resuscitation, use probiotics, milk opening time, interval between two stools gestational age main occurrence newborns, yielding importance scores 0.632, 0.407, 0.313, 0.258, respectively. A first time ≥ 24 h 3 days FI, while probiotics 34 weeks protective against newborns. practice, should improve perinatal care obstetrics aim reducing hypoxia delivery. When feeding, early opening, stimulation defecation measures be implemented FI.

Language: Английский

Citations

1
Fetal Health Classification Based on Tri-Ensemble Sampling DOI

凯 刘

Software Engineering and Applications, Journal Year: 2024, Volume and Issue: 13(02), P. 155 - 164

Published: Jan. 1, 2024

Language: Английский

Citations

0
CFCM-SMOTE: A Robust Fetal Health Classification to Improve Precision Modeling in Multiclass Scenarios DOI Creative Commons
Ahmad Ilham, Asdani Kindarto,

Akhmad Fathurohman

et al.

International Journal of Computing and Digital Systems, Journal Year: 2024, Volume and Issue: 15(1), P. 471 - 486

Published: April 29, 2024

The advent of cardiotocography (CTG) has radically transformed prenatal care, facilitating in-depth evaluations fetal health.Despite this, the reliability CTG is frequently undermined by data-related issues, such as outliers and class imbalanced data.To address these challenges, our study introduces an innovative integrated methodology that combines cluster-based fuzzy C-means (CFCM) with synthetic minority oversampling technique (SMOTE) to improve precision classification health status in multiclass scenarios.We used a considerable dataset from UCI Machine Learning Repository, employing CFCM manage SMOTE rectify data.This approach significantly improved performance algorithm, fact corroborated comprehensive experimental validation can be found Ref. [1].We observed notable improvements several evaluation metrics, including (PRC), sensitivity (SNS), specificity (SPC), F1 score (F1-S), accuracy (ACC), surpassing capabilities prior methodologies.Specifically, deployment algorithm amplified (PRC: 98.16% 99.58%), (SNS: 95.82% 100%), (SPC: 85.81% 99.75%), (F1-Score: 96.98% 99.79%), (ACC: 94.20% 99.84%) Classification Regression Tree (CART) for 'normal' class, while also improving Random Forest (RF) PRC: 94.77% 95.89% ACC: 90.60% 97.45%.These results confirm potential CFCM-SMOTE robust model diagnostics basic strategy development predictive analyzes healthcare.

Language: Английский

Citations

0
A Novel Active Learning Technique for Fetal Health Classification Based on Xgboost Classifier DOI
Kaushal Bhardwaj,

Niyati Goyal,

Bhavika Mittal

et al.

Published: Jan. 1, 2024

Ensuring safe pregnancy and reducing maternal infant mortality rates require addressing factors that affect fetal health. The application of machine learning algorithms in monitoring health helps to improve the chances timely intervention better outcomes case any possible issues fetuses. Existing studies offered aid this issue typically train models using a significant portion dataset, ranging mostly around 75%-80%. In work, we propose novel solution implementing an active technique identify most informative data samples for training model leading high accuracy with limited number samples. It employs query function built upon uncertainty diversity criteria which are derived based on properties XGBoost classifier. For deriving soft probabilities obtained unlabelled used while distance among feature space is utilized criteria. proposed approach shows superior performance comparison all state-of-the-art methods. Through analysis experimentation, achieves average higher than 99% by utilizing less 20% dataset training. This demonstrates its efficacy potential monitoring.

Language: Английский

Citations

0