Early Detection of Diabetes Using Machine Learning Algorithms and Performance Analysis

Abstract

Diabetes is a widespread metabolic disease and a major public health problem that affects millions of people worldwide. Early screening and risk prediction allow timely intervention for high-risk groups and significantly reduce the risk of severe complications. In recent years, machine learning (ML) has shown clear value in the analysis of large-scale medical data. This study reviews and compares mainstream models for diabetes prediction and proposes a new prediction framework. On the Pima Indians Diabetes Dataset, the proposed hybrid model reached an accuracy of 0.92, a precision of 0.91, a recall of 0.90, and an F1 score of 0.91, higher than the decision tree, support vector machine, random forest, and logistic regression models tested. Such a system may help medical staff with screening and clinical decision-making.