Reliable Fetal Health Assessment through Ensemble Techniques in Machine Learning

Ensuring the well-being of foetuses during pregnancy is crucial for preventing child and maternal mortality, aligning with the United Nations' Sustainable Development Goals. Utilizing a dataset of 2126 records from Cardiotocography (CTG) exams, obtained from the Kaggle repository, the study employs machine learning ensemble techniques, specifically stacking with 10x10 validations, to increase the classification of fetal health's precision and dependability. To address class imbalance, the dataset is resampled to 150%, and filter-based feature selection using the chi-squared statistical test is applied. Base models, including Decision Trees (C4.5), Random Forest (100 trees), Logistic Regression, and Sequential Minimal Optimization (SMO), are utilized. With the help of the Weka Machine Learning software tool, the ensemble model, after hyper parameter tuning and employing diverse techniques, with 98.68% accuracy, the model effectively classifies fetal health, indicating robust performance across various evaluation metrics.