An Explainable AI Framework for Early Diabetes Risk Prediction and Clinical Decision Support (Preprint)

BACKGROUND Diabetes mellitus is a major public health concern, and early identification of diabetes risk may support timely clinical assessment. Machine learning can identify patterns in demographic and clinical data, but the limited interpretability of some predictive models can make their outputs difficult for healthcare professionals to understand. OBJECTIVE This paper proposes an Explainable Artificial Intelligence (XAI) framework for early diabetes risk prediction that combines machine learning models with interpretable explanations to support clinical decision-making. METHODS The proposed framework includes data preprocessing, exploratory data analysis, feature selection, predictive modelling, and model evaluation. Logistic Regression, Decision Tree, Random Forest, and Gradient Boosting are considered as candidate classification algorithms. SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) are proposed to explain feature contributions at global and individual prediction levels. Planned evaluation metrics include accuracy, precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic Curve (ROC-AUC). RESULTS The framework is proposed but has not yet been experimentally evaluated. Therefore, no measured predictive performance, model comparison results, or SHAP and LIME explanation findings are available to report. CONCLUSIONS The proposed framework outlines an approach to combining diabetes risk prediction with interpretable model explanations. Experimental validation using an appropriate dataset is necessary to establish its predictive performance, generalisability, and potential usefulness in clinical decision support. The framework is intended to complement, not replace, professional medical judgement. CLINICALTRIAL Not applicable. This manuscript does not report a clinical trial.