Comparação do support vector machine com outros métodos de classificação: uma aplicação a dados de nascimentos prematuros

Abstract

Worldwide, thousands of births are premature, often causing irreversible damage to the health of the children and/or their mothers. This project aims to apply statistical techniques, more specifically a supervised machine learning method for binary classification (where the target variable can only assume two values, such as 1 if the baby is premature and 0 otherwise), to identify and quantify maternal and neonatal factors that may be associated with premature births. To this end, the Support Vector Machine (SVM) method will be used and its predictive performance will be compared with traditional parametric models, such as logistic regression, and with a semi-parametric model based on Gaussian processes.