Performance comparison machine learning algorithms in diabetes disease prediction

Abstract

Machine learning has been widely used in the field of medicine with the developing technology in recent years. Machine learning is a field that is also used in the diagnosis of diabetes and helps experts make decisions. Diabetes is a lifelong disease that is common worldwide and in our country. The main purpose of this study is to diagnose diabetes early using different machine learning classification algorithms. Another purpose of the study is to compare the success of the machine learning models used. Early diagnosis of diabetes allows to lead a healthy and normal life. In this context, it has been tried to diagnose diabetes early by using the machine learning techniques Decision Tree, Random Forests, K-Nearest Neighbor and Support Vector Machines classifiers on the Pima Indians Diabetes dataset. The dataset includes 9 features and 768 samples. Success evaluation of classifiers was made using Accuracy, Precision, Recall, F1-Score and AUC metrics. Random Forests gave the best results with 80 percent accuracy. This paper is to examine the association of different machine learning techniques usage, diabetes data diagnostic capabilities, diagnosis of diabetes in women diabetes patients and comparison of performances for machine learning techniques. Implications for theory and practice have been discussed. In this study, comparisons were made using different algorithms from the classification algorithms used in the literature and contributed to the literature in this field.