Skip to main content
CISTI'2023 - 18th Iberian Conference on Information Systems and Technologies

Full Program »

Machine-Learning Model For Classification of The Prognosis of Tuberculosis Using Real Data From Brazil

Tuberculosis (TB) was for many years, until the arrival of COVID-19, the world’s leading cause of death from an infectious agent. Despite efforts by the World Health Organization (WHO) to reduce the incidence of TB, it is estimated that in 2020 about 10 million people became ill with the disease and 1.3 million deaths were recorded worldwide as a result of the disease. In Brazil, one person contracts TB every five minutes and one person per hour dies from TB. In 2020 alone, there was an increase of 12% in the number of deaths recorded compared to 2019. Monitoring the possible outcomes of a patient with TB is an important task that can help reduce the early mortality of a patient diagnosed with this disease. However, determining this outcome is not a trivial task, especially in terms of anticipating the patient’s prognosis. Brazil has the and Notifiable Diseases Information System (in Portuguese, SINAN), which contains a database with the records of patients with notifiable diseases, including TB. Classifying the treatment outcomes of tuberculosis into either CURED and DIED (prognosis) using a machine learning (ML) model can assist healthcare professionals in deciding on the most appropriate treatment given the individual patient’s conditions and likely course of disease based on medical experience. In this article, we propose the use of ML to determine the prognosis of tuberculosis using support vector machines and gradient boosting from literature. We also feature selection techniques and random search techniques to find the hyperparameter optimization. Using a rigorous scientific methodology, experiments were carried out with different scenarios that balanced and imbalanced the data set and appropriate metrics were used to evaluate the models. Then, the model with the best performance was selected.

Maicon Herverton Lino Ferreira da Silva Barros
Universidade de Pernambuco (UPE)

Vanderson de Souza Sampaio
Instituto Todos pela Saúde (ITpS)

Patricia Takako Endo
Universidade de Pernambuco


Powered by OpenConf®
Copyright ©2002-2022 Zakon Group LLC