<p>The management of crises faces similar challenges, such as river overflows and devastating floods in mountainous areas and drought-prone areas, respectively. Accurate streamflow forecasting is important for the effective management of crises. The development of strong prediction models enhances streamflow prediction accuracy and reduces prediction time and hence is very imperative for successful crisis management. This study applied the eXtreme Gradient Boosting algorithm for machine learning in the prediction of streamflow. The meta heuristic optimization techniques are applied to improve the prediction accuracy of machine learning models, and it led to developing hybrid models with eight approaches to optimize the learning rate and the number of decision trees. The performances of these hybrid models have been assessed using various performance measures. These results of integration showed that XGBoost with the Aquila Optimization algorithm had high accuracy in comparison with other hybrid models, which decreased the time of prediction. The proposed hybrid model, integrating XGBoost and Aquila Optimization, demonstrated high efficiency with respect to R² values of 0.98 and 0.96 on the training and testing data set, respectively, and almost negligible overfitting. The study showcases the effectiveness of combining the XGBoost algorithm with meta-heuristic optimization techniques for streamflow prediction, particularly in sensitive management situations such as river overflows and flooding events.</p> Graphical abstract <p></p>