A PREDICTIVE ANALYSIS SYSTEM FOR TYPE 2 DIABETES USING DEEP LEARNING

Resumo

Abstract

Introduction: The World Health Organization has published its first global report on diabetes, concluding that from 1980 to the present day, the number of diabetics worldwide has quadrupled in a single generation. Diabetes, discovered thousands of years ago, has become a chronic disease of the 21st century, increasing the risk of stroke, kidney failure, peripheral vascular disease, heart disease, and death. Methodology: This study analyzed public diabetes data collected by the Centers for Disease Control and Prevention (CDC, USA) and developed a predictive model to classify the risk of type 2 diabetes using Deep Learning. The classification model was implemented using the Multilayer Perceptron Network (MLP) architecture. Results: The proposed model achieved an overall accuracy of 73.8%. When analyzing the Confusion Matrix, the accuracy obtained was 68.78% for the Normal class and 78.72% for the Diabetes class. Conclusions: The developed model was made available via web application, allowing users to enter data such as age, sex, body mass index (BMI), and smoking habits. The system then classifies the risk of type 2 diabetes and provides the probability of developing the disease, representing a useful tool for public health prevention strategies.

Keywords: Deep Learning; Machine Learning; Risk Prediction; Diabetes Mellitus, Type 2.