Análise e predição de mortalidade infantil utilizando modelos de aprendizado de máquina

Abstract

Infant mortality is a serious social problem that is still being discussed globally. Its rate is defined as the number of children who die before their first birthday for every 1,000 births. Given its urgency, the WHO (World Health Organization) has set a goal of ending preventable deaths of newborns by the age of 5 by 2030. Despite the reduction in this indicator, many countries still suffer from its occurrence. With advances in technology, it is possible to use machine learning to make predictions about this problem. The aim of this work is to study the SIM and SINASC data sets made available by DATASUS, thus carrying out a general analysis of the data using statistical methods to identify the main variables linked to infant mortality. Machine learning models are also developed to predict infant mortality. As the data is unbalanced, two resampling techniques were tested: SMOTE and Random Undersampling.