Diagnosing Diabetes with Machine Learning Techiques

Abstract

The rate of diabetes is rapidly increasing worldwide. Early detection of diabetes can help prevent or delay the onset of diabetes by initiating lifestyle changes and taking appropriate preventive measures. Until now, prediabetes and type 2 diabetes have proved to be early detection problems. There is a need for easy, rapid, and accurate diagnostic tools for the early diagnosis of diabetes in this context. Machine learning algorithms can help diagnose diseases early. Numerous studies are being conducted to improve the speed, performance, reliability, and accuracy of diagnosing with these methods for a particular disease. This study aims to predict whether a patient has diabetes based on diagnostic measurements in a dataset from the National Institute of Diabetes and Digestive and Kidney Diseases. Eight different variables belonging to the patients were selected as the input variable, and it was estimated whether the patient had diabetes or not. Of the 768 records examined, 500 (65.1%) were healthy, and 268 (34.9%) had diabetes. Ten different machine learning algorithms have been applied to predict diabetic status. The most successful method was the Random Forest algorithm with 90.1% accuracy. Accuracy percentages of other algorithms are also between 89% and 81%. This study describes a highly accurate machine learning prediction tool for finding patients with diabetes. The model identified in the study may be helpful for early diabetes diagnosis.