Abstract Subsurface uncertainty poses a serious difficulty in deploying geo-energy applications, owing to its complexity and our limited access to it. Reducing such uncertainty is essential to enhance the reliability of simulation results that define safe operating conditions. Ground deformation analysis is capable of contributing to reducing subsurface uncertainty. For example, a double-lobe ground deformation shape revealed a vertical fault zone at depth in the CO2 storage project at In Salah, Algeria. The aim of this work is to outline a process for reducing subsurface uncertainty by correlating subsurface characteristics with ground deformation data. The workflow begins with training of a supervised gradient boosting-based machine learning regression model that predicts ground deformation caused by reservoir pressurization.  We utilize a verified analytical solution (Wu, Rutqvist, and Vilarrasa, 2024) to assess ground displacement in response to pressurization of a reservoir intersected by either an impermeable or permeable fault to train the machine learning model. The instantaneous solution provided by the analytical solution enables us to generate an extensive dataset for training the model, encompassing fault and reservoir geometry as well as mechanical properties and operation conditions, i.e., reservoir pressurization. Simultaneously, principal component analysis and a simplified parametric space analysis are also performed. The results indicate that the pore pressure buildup and reservoir depth have the most significant impact on ground displacement. This study highlights that an appropriately trained machine learning model can effectively predict ground deformation and provide valuable information about the corresponding subsurface characteristics.