ARTIFICIAL INTELLIGENCE APPLIED TO HEALTH - DATA ANALYSIS FOR CHILDREN'S HEALTH

Abstract

Machine Learning (ML) is a subset of Artificial Intelligence plays an important role in healthcare, providing predictive models created from algorithms and large databases. These models can classify patients for diagnostic or prognostic purposes in various diseases. This research aimed to develop a predictive model for death due to Severe Acute Respiratory Syndrome (SARS) for children aged 0 to 3 years in the North region of Brazil, using data provided by the Brazilian Ministry of Health. An applied research was carried out using the CRISP-DM methodology that guided the entire process of selection, processing, transformation, application of ML algorithms and evaluation of the model. The Random Forest, Logistic Regression, K-Nearest Neighbors and XGBoost algorithms were used through the Weka software, where the model with Random Forest had superior performance. The model was generated with cross-validation and evaluated according to the metrics of sensitivity, specificity, accuracy, precision, F1-Score and AUC-ROC, the latter being the primary evaluation metric. Finally, a software application prototype for using the model was developed in the Java language so that the knowledge generated by the model reaches healthcare professionals.