Bridge piers in hydraulic environments are vulnerable to local scour. Water flow causes erosion of adjacent sediment vicinity of the piers leads a risk to bridge structure. This results in unique flow patterns near the bridge piers. Soft computing models that are data-driven and efficient are used by researchers nowadays to predict local scour. A total of 180 data sets from the previous literature over four decades were collected. Ensemble frameworks, such as Random Forest Regression (RF) and Extreme Gradient Boosting (XGBoost), are utilized in this study to compute the maximum local scour depth in clear water flow conditions. As independent variables, there are five input parameters: flow strength, flow shallowness, sediment gradation, sediment coarseness, and dimensionless time were considered. A total of three performance indicators were used namely Coefficient of Determination (R2), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE) to assess the effectiveness of these two machine learning models. Three literature equations with high potential have been taken into account for comparison with the current machine learning model about performance indicators. The results of the study show that the XGBoost performs better during the training stage with R2 = 0.983, MAE = 0.066, and RMSE = 0.086, while the RF performs best during the testing scenario with R2 = 0.765, MAE = 0.221 and RMSE = 0.299, The features importance found using XGBoost which indicates that flow intensity to be the most important variable that causes this phenomenon most vulnerable.