<p>Acute respiratory infection (ARI) remains one of the leading causes of morbidity and mortality among children under five in Bangladesh. Identifying key determinants of ARI using advanced data-driven approaches can enhance early detection and prevention. This study applied several Machine Learning (ML) algorithms to the 2022 Bangladesh Demographic and Health Survey (BDHS) data to predict ARI occurrence and identify its major socio-demographic and environmental predictors. A total of 5,184 child records were analyzed after data cleaning and preprocessing. Class imbalance was corrected using the Synthetic Minority Oversampling Technique (SMOTE). The Boruta algorithm was employed to select the most influential predictors. Five ML algorithms—Random Forest, Decision Tree, Extreme Gradient Boosting (XGBoost), Naïve Bayes, and Logistic Regression—were trained using an 80:20 train–test split with 5-fold cross-validation. Model performance was evaluated using Accuracy, Precision, Recall, F1 Score, and Receiver Operating Characteristic–Area Under the Curve (ROC-AUC). The Random Forest model achieved the best overall performance (Accuracy = 84.7%, Recall = 88.7%, F1 = 0.853, ROC-AUC = 0.93), followed by Decision Tree (AUC = 0.94) and XGBoost (AUC = 0.92). Feature importance analysis based on the Mean Decrease in Gini criterion revealed Geographic Division (0.186), Maternal Education (0.154), and Child’s Age (0.132) as the strongest predictors. Socio-economic status, sanitation, and media exposure also contributed meaningfully to ARI prediction. Tree-based ensemble models, particularly Random Forest, effectively captured nonlinear and interactive effects among predictors, outperforming classical models. These findings suggest that ML models can provide valuable decision-support tools to reduce ARI-related disease burden.</p>