Early Detection of Diabetes through Interpretable Machine Learning Models

Abstract

Early detection of diabetes is crucial for effective management and prevention of complications. Interpretable machine learning models offer a promising solution by leveraging large datasets to identify subtle patterns indicative of early diabetes onset. This study employs models such as decision trees, logistic regression, and explainable boosting machines (EBMs) to ensure transparency and interpretability, which are vital for healthcare providers' trust. Using clinical and demographic data, we trained and validated these models, utilizing tools like SHAP and LIME to highlight key predictors, including age, BMI, blood pressure, and family history. The models achieved high accuracy in predicting diabetes risk, providing actionable insights into contributing factors. Integrating these models into routine screening can enhance early detection rates, enabling timely interventions and reducing the burden of diabetes on individuals and healthcare systems, demonstrating the transformative potential of interpretable machine learning in chronic disease detection.