When constructing a dam in a mined-out area, it is necessary to handle the mined-out area effectively. Analyzing the distribution of mined-out areas is essential to ensure the safety of dam structures. In response to this issue, a machine learning-based method for analyzing the settlement of reservoir dam structures in coal mining subsidence areas is proposed. The method includes the following steps: first, analyze the geological conditions of the mined-out area and calculate the conditions for each depth of the mined-out area. Secondly, analyze the deformation mechanism of the mined-out area and calculate the impact of the mined-out area at representative depths. Next, use machine learning methods to regressively fit the deformation displacement and obtain the deformation function. Finally, validate the predicted depth based on on-site monitoring data and drilling information. The study concludes that deeper mined-out areas lead to larger total displacement, and as the depth increases, the displacement caused by self-weight stress gradually increases. Additionally, the displacement response of the dam load decreases with the increase in depth. The predicted depths align well with the actual depths. The proposed method is reasonable and feasible, providing a basis for reinforcement schemes in mined-out areas.