Balikpapan Bay contains extensive mangrove forests which play an important role as habitat for a range of species and in providing a range of ecosystem services. In recent years, the mangrove forests around Balikpapan Bay are increasingly being lost and degraded due to development pressures. Thus, change detection in mangrove ecosystem has become highly relevant, as it can provide essential information to support the conservation practices and coastal management. This study aims to map mangrove forest change in Balikpapan Bay, East Kalimantan over a five-year period from Sentinel-2 using machine learning. Five machine learning algorithms (Random Forest (RF), Support Vector Machine (SVM), Classification and Regression Tree (CART), K-Nearest Neighbors (KNN), and Minimum Distance), implemented on the Google Earth Engine platform, were evaluated to determine the most suitable method. The evaluation results indicate that RF, SVM, and CART yielded mangrove mapping accuracies of 80% or higher. Notably, the CART algorithm surpassed the other tested models, demonstrating the highest overall accuracy of 84% and a Kappa coefficient of 0.78. Mapping using the selected CART model shows that, between 2020 and 2025, mangrove areas in Balikpapan Bay decreased by 21% (2,906.17 ha). Approximately 97% (2,834.49 ha) of this loss is concentrated in the North Penajam Paser, which has a high rate of land conversion to built-up areas.