<p>The number of variables, including the properties of soil, groundwater conditions, and the construction technique used, affect the embedment depth (ED) and maximum bending moment (MBM) of cantilever sheet piles. The goal of this study was to predict the ED and MBM of cantilever sheet pile walls embedded in clay with sand as a backfill. Ridge regression, extreme gradient boosting, and polynomial regression (PR) are three machine learning (ML) approaches that were used for this purpose. Four key input parameters namely unit weight of sand, angle of shearing resistance of sand, unit weight of clay, and cohesion are taken into consideration while applying these three ML models to 400 datasets in order to predict ED and MBM. Statistical performance parameters like coefficient of determination (R²), variance account factor, Willmott’s index of agreement, root mean square error (RMSE), mean absolute error and weighted mean absolute percentage error were used to check the performance of the ML model. Additionally rank analysis, reliability index, scatter plot, Taylor diagram, comparative measure analysis, residual plot, and external validation were also used to evaluate the model performance. Based on numerous assessment indicators, the findings shows that PR performed better than the other suggested ML models in both scenarios, i.e., ED and MBM prediction. PR performs better in both the training (TR) and testing (TS) phases because it predicts ED with the highest R² (TR = 0.999 and TS = 0.997) and the lowest RMSE (TR = 0.003 and TS = 0.009), and it predicts MBM of cantilever sheet pile walls with the highest R² (TR = 0.999 and TS = 0.998) and the lowest RMSE (TR = 0.001 and TS = 0.002). The efficacy and resilience of the PR model as a useful tool for predicting the ED and MBM of cantilever sheet piles were shown by detailed investigation.</p>