Diabetes Mellitus Disease Prediction Using Logistic Regression (LR) and Support Vector Machine (SVM) Methods

Abstract

Diabetes Mellitus (DM), also known as diabetes or sugar disease, marked by high blood sugar levels and poses a major health issue in Indonesia with the number of cases increasing every year. Often referred to as the silent killer, DM often goes unnoticed due to its subtle symptoms, increasing the risk of severe complications if not treated promptly. The lack of information or awareness about the early symptoms of DM, limited time and cost in conducting health checks, and limited access to health services are challenges in detecting DM disease early. To overcome this problem, the development of a prediction model is essential to prevent serious complications. This study aims to create a predictive model using LR and SVM methods based on parameters such as pregnancy, glucose levels, blood pressure, skin thickness, insulin, BMI, diabetes pedigree, age, and outcome. The dataset used is DM disease risk data collected by Kaggle from the National Institute of Diabetes and Disgetive and Kidney Disease (NIDDK). Based on the research results, the LR method shows a better level of accuracy compared to the SVM method. The accuracy of the model using the Logistic Regression method is 79.31% while the SVM method has an accuracy value of 77.24%, with a difference in accuracy of 2.07%. This research applies hyperparameter tuning with Grid Search to find the best combination of hyperparameter.