Accurate prediction of structural displacement demands is a critical aspect of earthquake engineering for design and performance assessment. In addition to traditional methods available in the literature for estimating structural displacement demands under seismic excitations, energy-based approaches have gained increasing attention in recent years. This study investigates the effects of hyperparameters in the XGBoost machine learning algorithm (such as learning rate, maximum depth, number of estimators, and subsampling ratio) on the prediction performance of the relationship between seismic input energy and peak displacement demands of structures. Time history analyses were conducted on equivalent single-degree-of-freedom systems using ground motions recorded during the February 6, 2023, Kahramanmaraş Earthquakes. The sensitivity of the algorithm was evaluated for different hyperparameter combinations. Algorithm performance was assessed based on the accuracy of specific hyperparameter settings, generalization capability, and computational efficiency. The findings provide guidance on determining optimal hyperparameter settings for XGBoost, contributing to the improvement of energy-based displacement demand prediction models.