<p>Micropiles are widely recognized in geotechnical engineering for their effectiveness in bearing substantial loads, even in soft cohesive soils. However, accurate assessment of their load-carrying capacity remains limited, particularly in group configurations within these soils. Considering this, the current study specifically examines the prediction of the bearing capacity of micropiled raft foundations under defined settlement conditions in soft clayey soil. A new machine learning-based approach has been developed, employing a range of boosting algorithms to enhance the predictive accuracy of bearing capacity forecasting. Integrating the employed boosting methods, the proposed methodology aims to provide reliable estimations to improve foundation design in clayey soil conditions. To this end, different ensemble boosting methods, namely adaptive boosting, gradient boosting, extreme gradient boosting, and light gradient boosting machine, were implemented to forecast the load-bearing capacity of the micropiled raft using statistically significant features. This study utilized a dataset of 700 data points, each containing 6 distinct bearing attributes of cast-in-situ micropiled rafts subjected to static vertical load tests, to develop and validate the bearing capacity forecasting model. The assessment of the proposed forecasting approach demonstrated that the extreme gradient boosting algorithm provided highly accurate and robust predictions, highlighting its effectiveness for forecasting the load-bearing capacity of micropiled raft foundations. Furthermore, the efficiency of the extreme gradient boosting algorithm was validated using 150 laboratory experimental data points, and an additional set of 50 field data points collected from practical studies was employed for field validation.</p>