Accurate prediction of concrete compressive strength is essential for designing durable and cost-effective structural systems. This study investigates the use of ensemble learning models Gradient Boosting, AdaBoost, and XGBoost for predicting compressive strength based on mix-design parameters such as cement, water, superplasticizer, and silica fume content. A dataset comprising 649 concrete samples was preprocessed, normalized, and analyzed using correlation metrics to identify key predictors. The models were evaluated using $R^{2}$, RMSE, MAE, and MSE under 5-fold cross-validation. Results show that XGBoost outperforms the other models, achieving an $R^{2}$ of 0.8954 and RMSE of 6.40 MPa, followed by Gradient Boosting and AdaBoost. Residual analysis confirmed the superior stability and generalization of XGBoost. These findings demonstrate the potential of gradient-boosting algorithms for accurate and interpretable modeling of concrete strength, supporting their integration into intelligent material design workflows.