Diabetes Early Diagnosis System Using Machine Learning and Prognostic Feature Analysis

Abstract

Diabetes mellitus is a chronic metabolic disorder characterized by persistent hyperglycemia resulting from insulin deficiency or insulin resistance, and it has become a major global health concern due to its rapidly increasing prevalence. Since diabetes often remains asymptomatic in its early stages, late diagnosis leads to increased morbidity, mortality, and healthcare burden, making early detection essential. Recent advances in machine learning have enabled effective early disease diagnosis by analyzing large-scale medical data and identifying complex patterns beyond traditional methods. This paper presents a machine learning-based Diabetes Early Diagnosis System (DEDS) that utilizes prognostic feature analysis. The proposed system integrates data preprocessing, dimensionality reduction, feature selection, and multiple classification models to achieve accurate early-stage diabetes prediction. The PIMA Indian Diabetes Dataset is used for evaluation, incorporating clinical features such as glucose concentration, blood pressure, insulin level, body mass index, age, and hereditary factors. Principal Component Analysis (PCA) and Sparse PCA are applied to eliminate redundant features, improving model generalization and interpretability. Experimental results demonstrate that the proposed system outperforms traditional classifiers in terms of accuracy, sensitivity, and specificity, highlighting its effectiveness for early diabetes detection and its potential application in real-world clinical and telemedicine-based healthcare systems.