Assessment of structural health after earthquakes is a vital task, especially for high-rise buildings. This task is often time and resource-consuming. This study aims to train a machine learning (ML) model for the rapid prediction of the seismic structural response of high-rise steel moment frame buildings. Furthermore, only the acceleration response spectrum was considered ground motion input features instead of various intensity measures. The employed database considered buildings that were built with multiple geometric layouts and applied loads, as outlined in the current design guide. Nonlinear finite element analyses were conducted to obtain their responses under 240 ground motions. After the ML model was trained, the Shapley additive explanations (SHAP) method was then used to inspect the importance of input features. Spectral acceleration features were found to have a significant influence on the prediction. A reduced set of features was selected to retrain the model. The performance of the retrained model confirmed the importance of the acceleration response spectrum to the training of the model. Finally, the developed ML model managed to predict the structural response of the high-rise steel moment frame buildings, which ensured the successful application of machine learning for predicting the structural response under seismic situations.