Abstract

Diabetes is a chronic disease whose prevalence continues to rise globally and has become one of the leading causes of death according to the World Health Organization (WHO). It is characterized by high blood sugar levels caused by insulin production or function issues. In Indonesia, the increasing number of diabetes cases is not matched by public awareness of early detection, while limitations in medical personnel and examination time pose serious obstacles to screening. This issue encourages the use of machine learning technology as a solution to detect diabetes risk more quickly and accurately. The research aims to evaluate and identify the best algorithm and implement it into a web-based application using Streamlit, which both medical professionals and the general public can use. This study compares the performance of three machine learning algorithms—Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF)—in predicting diabetes risk based on clinical data. The study results show that the RF algorithm performed best with 97.03% accuracy and Area Under the Curve (AUC) 84.81%, followed by SVM and LR. The best-performing model was then integrated into the web application as an early diagnosis tool, which is expected to improve the efficiency, accuracy, and accessibility of early diabetes risk detection.

Concepts :

Access to Document

SDGs

Citations by Year