Enhancing Diabetes Prediction with Explainable AI Techniques

Abstract

Diabetes continues to be a significant global health challenge, necessitating accurate and timely prediction methods for early intervention. This abstract explores the application of explainable artificial intelligence (AI) techniques to enhance diabetes prediction models. Traditional machine learning models like logistic regression and decision trees are effective but often lack transparency in decision-making. In contrast, explainable AI techniques such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) offer insights into model predictions, elucidating which features contribute most significantly to diabetes risk assessment. By integrating these techniques with datasets such as the Pima Indians Diabetes Dataset and Electronic Health Records (EHRs), this study demonstrates improved interpretability and accuracy in predicting diabetes onset. Such advancements empower healthcare providers to make informed decisions, tailor interventions, and improve patient outcomes. The future of diabetes prediction lies in leveraging these explainable AI techniques to develop robust, transparent models that support proactive healthcare management and personalized patient care.