EXPLORANDO A APLICAÇÃO DE MODELOS DE APRENDIZAGEM DE MÁQUINA NA ANÁLISE DE SUSPEITA DE DIABETES

Uma Investigação Preliminar

Abstract

The conducted study addresses the possibility of employig machine learning models to assist in the early diagnosis of diabetes mellitus, a chronic condition that significantly impacts patients' quality of life and strains healthcare systems. The main objective is to explore the use of machine learning to aid in diabetes risk analysis, complementing traditional medical analysis. The dataset used was processed and normalized, and balancing techniques such as SMOTE and undersampling were employed, preparing the data to train three models: Keras Neural Network, Random Forest, and Gradient Boosting. Results show that the Random Forest model performs best overall, with high accuracy and the ability to minimize false positives, which is crucial given the study's context to prevent incorrect diagnoses or actual disease cases from going unnoticed. The study also highlights that synthetic data generation techniques can enhance the  representativeness of imbalanced medical datasets, reinforcing their potential for future  applications in medicine.