Pediatric respiratory disorders continue to be a major global health concern in the age of sophisticated computational frameworks, resulting in an increase in child death rate. This research improves mortality prediction in pediatric respiratory illnesses by using a strong machine learning framework and an extensive dataset from the Children’s Hospital of Rabat, Morocco. The following seven machine learning algorithms were assessed after extensive preprocessing, which included encoding, feature selection, oversampling, and hyperparameter tuning: Random Forest, K-Nearest Neighbors, Decision Trees, CatBoost, XGBoost, Gradient Boosting, and AdaBoost. With 98.84% accuracy, 92.66% precision, 93.79% recall, and 92.90% F1-score, gradient boosting fared better than the others. Explainable AI techniques such as SHAP guaranteed clinician trust and model interpretability.