Boletim COVID-PA: relatos sobre projeções baseadas em inteligência artificial no enfrentamento da pandemia de COVID-19 no estado do Pará

Gilberto Nerino de  Souza Jr.; Marcus de Barros  Braga; Luana Lorena Silva  Rodrigues; Rafael da Silva  Fernandes; Rommel Thiago Jucá  Ramos; Adriana Ribeiro  Carneiro; Silvana Rossy de  Brito; Cícero Jorge Fonseca  Dolácio; Ivaldo da Silva  Tavares Jr.; Fernando Napoleão  Noronha; Raphael Rodrigues  Pinheiro; Hugo Alex Carneiro  Diniz; Marcel do Nascimento  Botelho; Antonio Carlos Rosário  Vallinoto; Jonas Elias Castro da  Rocha

doi:10.1590/s1679-49742021000400012

##article.authors##

Gilberto Nerino de Souza Jr. Universidade Federal Rural da Amazônia https://orcid.org/0000-0002-8452-3938
Marcus de Barros Braga Universidade Federal Rural da Amazônia https://orcid.org/0000-0003-1621-4460
Luana Lorena Silva Rodrigues Universidade Federal do Oeste do Pará https://orcid.org/0000-0001-9480-7689
Rafael da Silva Fernandes Universidade Federal Rural da Amazônia https://orcid.org/0000-0002-3035-8025
Rommel Thiago Jucá Ramos Universidade Federal do Pará
Adriana Ribeiro Carneiro Universidade Federal do Pará
Silvana Rossy de Brito Universidade Federal Rural da Amazônia
Cícero Jorge Fonseca Dolácio Universidade Federal do Paraná
Ivaldo da Silva Tavares Jr. Universidade Federal de Viçosa
Fernando Napoleão Noronha Universidade Federal Rural da Amazônia
Raphael Rodrigues Pinheiro Universidade Federal Rural da Amazônia https://orcid.org/0000-0002-2457-7571
Hugo Alex Carneiro Diniz Universidade Federal do Oeste do Pará
Marcel do Nascimento Botelho Universidade Federal Rural da Amazônia
Antonio Carlos Rosário Vallinoto Universidade Federal do Pará
Jonas Elias Castro da Rocha Universidade Federal Rural da Amazônia https://orcid.org/0000-0002-0255-087X

DOI:

https://doi.org/10.1590/s1679-49742021000400012

Keywords:

COVID-19, Artificial Intelligence, Forecasting, Neural Networks, Decision Making

Abstract

Objective: Report the university research and extension product denominated ‘Boletim COVID-PA’ which presented projections about the pandemic in the State of Pará, Brazil, with practical, mathematically rigorous and computationally efficient approaches. Methods: The artificial intelligence technique known as Artificial Neural Networks was used to generate thirteen bulletins with short-term projections based on historical data from the State Department of Public Health system. Results: After eight months of projections, the technique generated reliable results with an average accuracy of 97% (147 days observed) for confirmed cases, 96% (161 observed days) for deaths and 86% (72 days observed) for occupancy of intensive care unit beds. Conclusion: These bulletins have become a useful tool for decision making by public managers, assisting in reallocating hospital resources and optimizing COVID-19 control strategies for the various regions of the State of Pará.