Diabetes Prediction Using Unsupervised Learning

Abstract

Diabetes is a chronic disease that affects millions of people worldwide. Diabetes complications can be avoided, and a patient's quality of life can be considerably improved if diabetes is detected and diagnosed early. Serious complications, such as heart disease, renal failure, and blindness, may not occur as a result. This paper presents research on the use of unsupervised learning algorithms for diabetes prediction. The dataset utilized in this study is made up of patient medical information from patients with and without diabetes. We use clustering and anomaly detection algorithms to uncover patterns and abnormalities in the data, and then we use these patterns to predict the risk of diabetes in new patients. The proposed method aims to identify patient subgroups based on clinical and demographic similarities, which can aid in the early detection of diabetes and customized medication. Using a variety of criteria, we examine and evaluate the performance of several unsupervised learning methods.