Driven piles are a common geotechnical solution for foundations on weak soil profiles. However, the hammer impacts during the driving process can generate excessive ground vibration levels, which, in extreme cases, can affect nearby structures and people. The complexity of this phenomenon often requires advanced numerical modelling, which entails high computational costs and time requirements. This work presents the development, training, and validation of a surrogate model based on machine learning model. A group of popular algorithms will be used, such as Artificial Neural Networks (ANNs), Multiple Linear Regression, Regression Tree, Random Forest, Gaussian Process Regression, Support Vector Machine, and Extreme Gradient Boosting (XGBoost). The models were trained to predict ground surface vibration levels in the spatial and frequency domain. The models were trained using an extensive database generated from a numerical model, previously developed. The study shows the contrast between the algorithms in understanding the data pattern. ANNs and XGBoost show notable understanding of the data trending in frequency and spatial domain over other algorithms. This places a spotlight on the abilities of machine learning algorithms to address the geotechnical problems driven by high complexity levels.