<p>This paper focuses on the prediction of the amplitude and the force transferred to supporting soils due to vertical vibrations of a massive foundation on transversely isotropic poroelastic soils using machine learning models. Five ML models, namely artificial neural network (ANN), long short-term memory (LSTM), random forest (RF), extreme gradient boosting (XGBoost), and symbolic regression (SR), were established. These models were trained using a large volume of high-fidelity data derived from semi-analytical solutions and discretization techniques grounded in poroelastic theory. All developed ML models were capable of analyzing the saturated soils (either fully permeable or impermeable foundation) and the dry soils. The measurement of all models’ performance indicates that the output variables were well explained through the soil properties, the frequency, and the foundation mass, with <i>R</i>² exceeding 97% across all developed models. Additionally, the equations for amplitude and force transferred were explicitly obtained from the SR model. Results from all developed models were subsequently verified with various solutions to demonstrate their accuracy, with the ANN solutions displaying the best accuracy among all ML models. Besides, a sensitivity analysis identified the excitation frequency as the most significant parameter, with the first-order sensitivity of more than 80% in predicting the amplitude and force transferred to soils from vertical vibrations of a massive foundation across all soil types. Additionally, the developed models were validated using 1,800 additional data points to assess their predictive accuracy and generalizability in capturing the complex behavior of poroelastic soils under realistic loading and fluid pressure conditions.</p>