Comparative performance analysis of ensemble learning methods for fetal health classification

Fetal health monitoring is vital for early diagnosis and intervention during pregnancy, with cardiotocography (CTG) being a standard tool for assessing fetal well-being. However, CTG interpretation often suffers from subjectivity and inconsistency, motivating the need for automated, accurate, and interpretable diagnostic models. To address these challenges, we propose a robust machine learning framework that combines effective feature selection and ensemble learning with explainability. Specifically, Recursive Feature Elimination is used to reduce redundancy and identify the ten most discriminative features. An ensemble classifier, integrating Decision Tree, Random Forest, and Gradient Boosting, is developed to enhance classification accuracy. The model is trained and evaluated on a publicly available CTG dataset, achieving 99.56% accuracy, 99.54% precision, 99.59% recall, and a 99.56% F1 score. To ensure generalization, we conducted K-fold cross-validation and confusion matrix analysis. For interpretability, the framework incorporates Local Interpretable Model-agnostic Explanation, revealing influential features in each prediction. Compared to existing approaches, our model demonstrates superior performance and transparency, offering a practical and reliable decision-support system for fetal health assessment in clinical settings.