The contamination of water bodies in the Peruvian Amazon, particularly in the Madre de Dios region, has increased significantly due to illegal mining activities that severely impact human health and ecosystems. This issue is exacerbated by the lack of effective tools for monitoring and managing water bodies, which could help mitigate negative effects and ensure their preservation. In this study, water bodies were identified using machine learning and satellite image analysis from the area known as 'La Pampa', a zone severely affected by illegal mining located between kilometers 98 and 115 of the Interoceanic Highway in the Madre de Dios region, Peru. Using Google Earth Engine, 600 satellite images were collected and classified into three categories: water bodies (200), soil (200), and vegetation (200). Models such as Random Forest, K-Nearest Neighbors, Support Vector Machines, and a Multilayer Perceptron were trained and validated. The results show that the K-Nearest Neighbors model achieved the best performance, with a precision of 92.58%, recall of 92.74%, F1-Score of 92.61%, and an accuracy of 92.49%, outperforming the other evaluated models. These findings highlight the feasibility of combining machine learning with satellite images for the management of water resources in affected areas, offering a valuable tool for environmental decision-making.