The pressuremeter test (PMT), a valuable geotechnical in situ test, is used to design foundations of varying depths (shallow, semi-deep, and deep). It assesses a soil's bearing capacity and settlement through two key parameters: limit pressure and pressuremeter modulus. However, the high cost and time demands of PMTs limit their widespread use. This study addresses this challenge by exploring the effectiveness of ensemble machine learning algorithms. To achieve this main goal, we employ two methods, Extreme Gradient Boosting (XGBoost) and Random Forest, to predict limit pressure of soil. To develop the mentioned models an experimental database was used to train and validate the developed models. The effectiveness of these methods are evaluated using three statistical metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R²). The performance metrics show that the developed XGBoost and Random Forest models are viable alternatives to the pressure meter test for estimating limit pressure. Both models achieved high R-squared values (around 0.99 for training and 0.90 for testing) and a low root mean squared error (RMSE) of 3.23 and 4.13 for the testing set, respectively. These results demonstrate the effectiveness of using machine learning in the geotechnical field. To further understand the influence of individual factors on the predictions, we will utilize the Shapley Additive explanations (SHAP) method. This technique analyzes the contribution of each input variable (feature) to the model's predictions of limit pressure. By quantifying the importance of these features; SHAP provides valuable insights into which soil properties most significantly affect the foundation design parameters.