Ml Caries Risk Stratification In Post-Anesthesia Pediatric Patients

To develop machine learning (ML) models predicting postoperative caries risk in children undergoing general anesthesia, comparing logistic regression, XGBoost, SVM, and random forest. Retrospective analysis of 200 children (1-5 years) treated under anesthesia (2023-2024). Data were split into 70% training and 30% validation sets. Four ML models were implemented in Python, evaluated via accuracy, F1-score, and AUC-ROC. XGBoost outperformed all other models across key metrics (p < 0.05). It achieved the highest accuracy (87.5%), F1-score (84.8%), and AUC (0.91), significantly exceeding SVM, RF, and LR. Although logistic regression showed a relatively high F1-score (84.1%), its overall accuracy was the lowest (85.2%), highlighting a discrepancy between precision-recall balance and classification accuracy. XGBoost demonstrates optimal caries risk prediction in anesthetized pediatric patients, with high discriminative power (AUC>0.9) and balanced precision-recall. This supports ML integration into preventive dental care protocols.