This study evaluates LightGBM, XGBoost, and a hybrid LIGB model for diabetes prediction using Kaggle data, with a focus on improving person-centric health care. The aim was to identify the best model for predicting diabetes while enhancing personalized health outcomes. LightGBM quite effectively categorized diabetic cases with 95.43% accuracy. With 91.63% precision and 91.25% recall, the model lowers false positives and correctly detects diabetes sufferers. With an F1 score of 87.86%, the precision and recall of the F1 exhibit well-balance. With 94.27% accuracy, XGBoost did nicely. This result indicates that it manages false positives and negatives rather well. The model achieves balanced categorization according to accuracy of 90.55% and recall of 91.62%, with 90.72% F1 score shows it can regulate several data trends. The hybrid LIGB model with XGBoost and LightGBM was 94.85% precise. Both models were used to create a classification system with 91.09% precision and 91.43% recall. This hybrid strategy improves forecast accuracy by controlling model bias and volatility. This study shows how hybrid machine learning can enhance diabetes prediction and other healthcare outcomes. This study demonstrates how hybrid machine learning techniques can significantly improve person-centric diabetes prediction and healthcare outcomes.