Detecção de diabetes em pacientes adultos usando inteligência artificial explicável

Abstract

Diabetes, a rapidly expanding chronic condition
worldwide, requires diagnostic methods that are both accurate
and efficient. This study presents an artificial intelligence-based
approach, utilizing a multilayer perceptron neural network
designed to identify diabetes based on multiple influential factors.
The network was trained using a database, achieving an 82% pre-
cision and 83% accuracy. Additionally, the explainable artificial
intelligence method SHAP was applied, revealing the significant
contribution of parameters such as age and high blood pressure
in the diagnostic process. These results highlight the importance
of early detection and proper treatment of diabetes, reinforcing
the potential of AI as a clinical support tool.