<p>The response of soft soil improved by stone columns against high-speed train passage was analyzed numerically using a finite element program. The train speed, the length and diameter of the stone columns, and the axial stiffness and length of the geogrid encasement were among the parameters that were examined. The random forest (RF), support vector regression (SVR), and extreme gradient boosting (XGBoost) algorithms were used to predict the settlement over time caused by high-speed train passage on soft soil improved with stone columns under different conditions. Integrating the advantages of physics-based simulation and data-driven prediction, machine learning and finite element numerical modeling significantly improved the accuracy, efficiency, and insight of complex soil-structure interactions under high-speed train loads. Numerical simulations showed that the use of stone columns with a length/diameter (L/D) ratio of 17.5 in soft soil for a train passing at a constant speed of 80 m/s resulted in a 69.4% decrease in residual settlement and a 91.8% decrease in differential settlement along the railway track. Moreover, a comparison of the machine learning predictions demonstrated that XGBoost and RF achieved higher accuracy than SVR, with <i>R</i>² values of 0.998 and 0.997 for the training dataset, respectively.</p>