Accurate prediction of the dynamic response of simply supported bridges under varying loading conditions (particularly earthquake) is crucial for ensuring their structural safety and longevity. This paper discusses the response of simply supported bridges with moving loads and investigates the application of machine learning (ML) techniques, specifically boosting algorithms, to predict bridge responses to both moving loads and seismic events. A finite element method (FEM) model is used to conduct dynamic analysis simulating train moving loads and seismic events on a simply supported bridge. The resultant responses were meticulously compiled into a dataset, serving as the foundation for constructing a surrogate model employing boosting algorithms. Subsequently, the performance of this boosting model was compared with that of the conventional linear regression approach. Results showed that the R2 values for XGBoost are close to 1, making it powerful for structured data and predictive modeling tasks. Additionally, the feature importance analysis reveals that PGA is the dominant feature for both displacement and acceleration predictions.