A REVIEW ON DIABETIC DISEASE PREDICTION USING MACHINE LEARNING IN HEALTHCARE SECTOR

Abstract

Diabetes is a long-term illness characterised by high blood sugar levels as a consequence of either a lack of insulin
production or an ineffective action of insulin. The current research outlines in detail the diverse types of diabetes, their symptoms,
associated complications, and main risk factors, while also mentioning machine learning (ML) methods that are proficient at
diagnosis and prognosis. Autoimmune destruction of β-cells leads to Type 1 diabetes, while Type 2 develops from insulin
resistance and ultimately results in β-cell death; both processes are accompanied by oxidative and reductive stress. The major
symptoms observed through clinical examinations, including eyesight issues, reduction of the body, difficulties with urination,
and delayed recovery of wounds, all indicate the disease's slow nature. A prolonged period of high blood sugar is a cause of severe
complications, which can be liver cirrhosis, NAFLD, NASH, and cases of liver disease due to alcohol. The paper also discusses
different ML methods, including Decision Trees, Random Forests, AdaBoost, XGBoost, K-means, DBSCAN, and Autoencoders,
both supervised and unsupervised, for the purposes of early patient detection, stratification, and risk prediction. By merging AI
with medical science, the study has shed light on the role that ML-based systems can play in diabetes diagnosis, enriching it,
supporting individualised treatment, and progressively reducing the global burden of diabetes mellitus.