Enhancing Fetal Health Monitoring through TPOT and Optuna in Machine Learning-Driven Prenatal Care
A. Akilandeswari, G Arasuraja, Nagendar Yamsani, S. Radhika, N. Legapriyadharshini, S. Padmakala
Abstract
This study delves into the application of advanced machine learning techniques for the classification of fetal health, a critical domain in prenatal care. Utilizing a dataset based on cardiotocograms (CTGs), which record key fetal indicators like heart rate and uterine contractions, we compare two distinct machine learning approaches: a Random Forest Classifier optimized with the hyper parameter tuning tool Optuna, and a genetic programming-based model developed using TPOT (Tree-based Pipeline Optimization Tool).The Random Forest Classifier, configured with specific hyper parameters, delivered an accuracy of 94.13% and an impressive AUC of 0.9826. In contrast, the TPOT-optimized model, a Gradient Boosting Classifier with finely tuned parameters, achieved a higher accuracy of 96.01% and an internal CV score of approximately 95.24%. This comparison underscores the strengths and potential applications of these advanced methodologies in predicting and ensuring fetal health.