An Insight of Diabetes Detection Performance using Machine Learning Techniques

Diabetes is a metabolic disorder that affects many people around the world. Every year, its incidence rates are rising rapidly. Diabetes-related problems in many cases have the potential to affect the body’s vital organs and become lethal if left untreated. As this disease remains in asymptotic condition for long duration so its early detection can help. So, it must be detected early in order to be treated in a timely manner and prevent further complications. This paper presents a comprehensive review of machine learning approaches for early detection of Type 2 Diabetes Mellitus (T2DM). The study addresses the critical need for improved diabetes prediction methods and proposes automated diagnosis systems using various machine learning techniques. Multiple algorithms like Logistic regression, support vector machine, Knearest neighbour and Random Forest were evaluated across different datasets, with Random Forest emerging as the most effective classifier, achieving up to 99% accuracy in some studies. The research examined diverse data sources including electronic health records and specialized datasets, focusing on key predictors such as fasting plasma glucose, HbA1c, and BMI. While the results demonstrate the transformative potential of AI-driven technologies in diabetes care, challenges remain regarding limited sample sizes and clinical validation. Future work aims to expand datasets, improve preprocessing methods, and explore applications in other diseases.