Predicting Neonatal Condition at Birth through Ensemble Learning Methods in Pregnancy Care

Prematurity represents the determinant cause of infant mortality. This serious public health problem is directly related to the assistance provided during pregnancy and childbirth. Hence, this paper proposes the use of leading machine learning techniques capable of supporting health experts in pattern recognition in the prediction of high-risk situations for the fetus. The proposed model creates an ensemble of nearest-neighbor classifiers using the random subspace algorithm, reaching an overall accuracy of 0.937 and area under the curve of 0.721, in predicting the Apgar score, and 0.829 and 0.669 in predicting if the newborn will be small for gestational age, respectively. These results show the model effectiveness in reducing severe pregnancy related-problems.