An Ensemble Machine Learning and LLM-Augmented Framework for Real-Time Diabetes Risk Prediction and Lifestyle Guidance

Abstract

Diabetes is a major global health concern, projected to affect more than 850 million people by 2050. Early risk prediction is crucial for timely intervention and improved patient outcomes. This study presents a machine learning framework for diabetes prediction using a clinical dataset, with class-weighted learning to address imbalance synthetic oversampling. Multiple models are evaluated, including LR, SVM, KNN, DT, RF, BRF, EE, and variants of Gradient Boosting, where RF achieved the best balance of accuracy and robustness. The model's reliability is ensured using SHAP interpretation. The best-performing model is deployed in a web application for real-time predictions and personalized lifestyle guidance to assist the user in real time. The result demonstrates the effectiveness of the ensemble models and the potential of combining predictive analysis with LLM for patient-centric health care solutions.

Ähnliche Arbeiten