Explainable and Interpretable Machine Learning Frameworks for Early Diabetes Risk Prediction and Clinical Decision Support

Abstract

Early detection of diabetes is essential for preventing severe health complications, improving patient outcomes, and enabling timely medical intervention through personalized healthcare strategies. With the growing availability of healthcare data and advancements in artificial intelligence, interpretable machine learning approaches have emerged as promising tools for supporting accurate and transparent diabetes risk prediction in clinical environments. This study investigates the application of explainable and interpretable machine learning techniques for early diabetes prediction while emphasizing model transparency, clinical reliability, and decision-making trustworthiness. Using the Pima Indians Diabetes Dataset, comprehensive data preprocessing procedures were performed, including missing value handling, feature normalization, outlier management, and selection of clinically relevant attributes such as glucose concentration, body mass index (BMI), insulin levels, age, blood pressure, and skin thickness measurementsVisualization-driven analyses were also employed to enhance the understanding of model behavior, feature importance, and patient-specific prediction explanations.