Multi-Domain Fusion with Explainable Boosted Learning (MF-EBL) for Diabetes Prediction

Diabetes is a chronic metabolic disorder that is rapidly increasing in the world, which makes early and accurate prediction critical for timely intervention. The study introduces a new predictive methodology called Multi-Domain Fusion with Explainable Boosted Learning (MF-EBL) for predicting diabetes using clinical, demographic, and lifestyle datasets. The multi-domain fusion inputs advanced methods of machine learning (ML) including XGBoost, Random Forest(RF), Support Vector Machine(SVM), Logistic Regression(LR), K-Nearest Neighbors (KNN), and Neural Networks that were evaluated according to accuracy, F1 score, AUC-ROC, and explainability or interpretability of the ML model. Overall, XGboost prediction model performed the highest in all methods studied (accuracy of 88.7%, F1-Score of 0.87 and AUC-ROC of 0.91). The SHAP is applied to understand the interpretability of the model to predict diabetes without losing model accuracy. The study `conducted a data analysis and feature selection for the all data included in the study to provide high confidence for a robust model, and gained insight into the key risk factors such as glucose levels, age and body mass index (BMI). The MF-EBL framework presents a robust and trustable solution for monitoring and early detection of diabetes with results that are meaningful to clinicians who are providing personalized care to patients and the results can assist with decision-making not only through monitoring and also with early risk assessment.