Application of ensemble learning in predicting shallow foundation settlement in cohesionless soil
Ningthoujam Jibanchand, Konsam Rambha Devi
Abstract
Due to significant uncertainties associated with soil, it is challenging to anticipate settlement accurately for shallow footings on cohesionless soil. To produce more precise predictive settlement models, four ensemble learning models have been created in this study: Bagging, Random Forest (RF), Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost). The models are created utilizing a sizable database based on standard penetration tests (SPT). A variety of evaluation criteria, including R2, RMSE, and MAE, were employed to rate the performance of the models. The analysis results showed that Bagging and XGBoost models demonstrate excellent performance with R2 values of 0.901 and 0.915, respectively, surpassing other models studied here as well as other models from the literature. Consequently, Bagging and XGBoost can be effective methods for predicting settlement in shallow foundations on cohesionless soil.