Use of Heuristic Regression Techniques in Stability Control of Cantilever Retaining Walls

This study investigates the stability conditions of cantilever retaining walls via heuristic regression techniques. The well-known heuristic regression techniques have been preferred to gain results quickly. In this study, heuristic regression techniques, including the M5 model tree and the Multivariate Adaptive Regression Splines, have been employed to investigate the model that provides sliding, overturning, and bearing capacity safety factors. Root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) have been utilized in the comparison of estimation for safety factors from the numerical analyses of 1024 different cantilever retaining wall designs. Reasonable values of RMSE, MAE, and R2 for models of safety factors have been obtained as 0.046, 0.008, 0.462 for sliding, 0.036, 0.007, 0.275 for overturning, and 0.999, 0.999, 0.998 for bearing capacity, respectively. This result demonstrates that an improved model by heuristic regression techniques for the cantilever retaining wall’s stability check can be employed reliably and effectively.