Comparison of Binary Logistic Regression and SVM to Classify Diabetes Sufferers

Abstract

Diabetes is a chronic metabolic disorder characterized by high levels of glucose in the blood due to disruption of the insulin hormone which functions as a regulator of the balance of blood sugar levels. This disease continues to increase in prevalence in various countries, making it a global health problem. Diabetes has trigger factors that contribute to the incidence of the disease, such as age, gender, smoking habits, healthy eating patterns, high blood pressure, and others. Diagnosis of diabetes can be done by carrying out a fasting blood sugar test, a 2-hour postprandial (PP) blood sugar test, and a random blood sugar test. However, it is very possible for diagnoses made by health workers to have errors due to subjectivity and different experiences, so a fast and precise classification method is needed to classify patients undergoing diabetes examination based on variables related to diabetes. The classification method used in this research is binary logistic regression and Support Vector Machine (SVM). A similar study carried out classification of diabetes sufferers using the Naive Bayes and KNN methods by comparing the results with SVM, so in this study the binary logistic regression method and SVM will be used to determine the performance of the classification method. The data used is secondary data. Next, the data is divided into training and testing data. The analysis results show that the SVM method is slightly superior in classification accuracy of testing data, namely 97.75%. With this research, it is hoped that decisions on patients undergoing diabetes examination will be faster, more precise and effective, and classification methods with better performance can be applied