Prediction of Apnea of Prematurity in neonates using Support Vector Machines and Random Forests

Machine Learning has a wide array of applications in the healthcare domain and has been used extensively for analyzing data. Apnea of Prematurity is a breathing disorder commonly observed in preterm infants. This paper compares the usage of Support Vector Machines and Random Forests, which are supervised learning algorithms, to predict Apnea of Prematurity at the end of the first week of the child's birth using data collected during the first three days of neonatal life. This paper also uses an optimization method called Synthesized Minority Oversampling Technique (SMOTE) to resolve the class imbalance problem observed in the data. Principal Component Analysis and one-hot encoding have been implemented for feature extraction and data preprocessing respectively. Among the results obtained, an AUC of 0.72 using the amalgamation of Random Forests and SMOTE is found to be the most accurate model.