PERBANDINGAN BERBAGAI MODEL MACHINE LEARNING UNTUK MENDETEKSI DIABETES

Abstract

Diabetes mellitus, often known as diabetes, is a significant metabolic illness that has a negative impact on living organisms. It causes high blood sugar levels by either creating inadequate insulin or using it inefficiently. Diabetes that is not effectively treated raises the risk of heart attacks, retinopathy, vision loss, skin disorders, and other ailments. Early detection is critical for guiding essential actions. In this setting, machine learning (ML) has emerged as a potent tool. We used Python data manipulation tools to develop ML techniques for discovering patterns and risk factors in the Pima Indian diabetes dataset in our study. We correctly identified patients as diabetes or non-diabetic using K-Nearest Neighbors (KNN), AdaBoost, Logistic Regression (LR), Light Gradient Boosting, Random Forest (RF), dan Support Vector Machine (SVM). Notably, we used the Synthetic Minority Over-sampling Technique (SMOTE) to solve class imbalance, which enhanced model performance. By efficiently utilizing ML and SMOTE in diabetes categorization, our work greatly adds to the scientific area. We suggest studying cutting-edge technology and undertaking external validation and clinical studies to assure trustworthy and generalized models for diabetic patient care in the future. With diabetes's increasing prevalence, such improvements have enormous promise for improving early identification and management, eventually leading to better health outcomes.