Performance Comparison of Machine Learning Models for Diabetes Prediction

Abstract

Diabetes is a chronic disease that causes blood sugar to rise. This chronic disease can be cause mortality. There are different diagnoses and treatment methods for diabetes in the medical field. In addition, with the developing technology, diagnosis of the disease can be made computer-aided. Computeraided diagnostic methods are a successful, fast, and alternative method that supports the doctor's decision. The use of computeraided diagnosis approach for diabetes and many other diseases is increasing day by day. Machine learning classification methods are the most commonly used methods for computer-aided diagnostics. The aim of this study is to design a model to detect the probability of diabetes in patients at an early stage with maximum accuracy. Therefore, seven machine learning classification algorithms were used, namely Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Gaussian Naive Bayes, Decision Tree, Random Forest, and Artificial Neural Network. The study was carried out on the Pima Indians Diabetes Database (PIDD) taken from the Kaggle database. The performances of machine learning methods were evaluated according to precision, recall, ROC curve, and PRC criteria. According to the results, the Logistic Regression method is more successful than other methods in classifying diabetes disease accurately.