Previsão da taxa de mortalidade infantil no Ceará: uma abordagem com aprendizado de máquina
Prediction of the Child Mortality Rate in the State of Ceará: A Machine Learning Approach
DOI:
https://doi.org/10.5380/re.v46i88.97156Keywords:
Taxa de mortalidade infantil, Previsão de séries temporais, Modelos de aprendizado de máquina, CearáAbstract
A Taxa de Mortalidade Infantil (TMI) é um indicador crucial do bem-estar e do desenvolvimento regional. O Ceará vem reduzindo a TMI ao longo do tempo. Nesse contexto, este trabalho objetiva prever a TMI cearense de 2001 a 2023 utilizando modelos Autoregressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), Decision Tree Regression (DTR), Random Forest Regression (RFR) e Extreme Gradient Boosting (XGB). O número de óbitos a cada mil nascidos vivos para tal período foi extraído do IPECEDATA. Considerou-se 75% da série para treinamento dos modelos e o restante para teste. Dentre esses, o XGB apresenta menor erro de previsão. Usando este algoritmo, constata-se, na projeção da TMI para 6 anos à frente, que esta taxa permanece nos mesmos patamares dos seis anos anteriores, com média de 11,5 óbitos a cada mil nascidos vivos. Portanto, recomendam-se políticas públicas no Ceará, visando reduzir a TMI no médio e longo prazos.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Manoel Alexandre de Lucena, Marcos Antonio Carvalho Alves de Sousa Júnior, Eliane Pinheiro de Sousa

This work is licensed under a Creative Commons Attribution 4.0 International License.
By submitting a manuscript to this journal, authors agree with the following terms:
1. Authors retain copyright and grant the journal the right of first publication, with the work simultaneously licensed under the Creative Commons Attribution CC-BY License, which allows for the sharing of the paper while acknowledging authorship and initial publication in this journal.
2. Authors are granted the right to enter into additional and separate contracts aiming at the non-exclusive distribution of the version of the manuscript published in this journal (e.g., reproduction either in institutional repositories or as book chapters), with the acknowledgment of authorship and initial publication in this journal.
