Novel Hybrid XGBoost Model to Forecast Soil Shear Strength Based on Some Soil Index Tests

Abstract

When building geotechnical constructions like retaining walls and dams is of interest, one of the most important factors to consider is the soil’s shear strength parameters. This study makes an effort to propose a novel predictive model of shear strength. The study implements an extreme gradient boosting (XGBoost) technique coupled with a powerful optimization algorithm, the salp swarm algorithm (SSA), to predict the shear strength of various soils. To do this, a database consisting of 152 sets of data is prepared where the shear strength (τ) of the soil is considered as the model output and some soil index tests (e.g., dry unit weight, water content, and plasticity index) are set as model inputs. The model is designed and tuned using both effective parameters of XGBoost and SSA, and the most accurate model is introduced in this study. The prediction performance of the SSA-XGBoost model is assessed based on the coefficient of determination (R2) and variance account for (VAF). Overall, the obtained values of R2 and VAF (0.977 and 0.849) and (97.714% and 84.936%) for training and testing sets, respectively, confirm the workability of the developed model in forecasting the soil shear strength. To investigate the model generalization, the prediction performance of the model is tested for another 30 sets of data (validation data). The validation results (e.g., R2 of 0.805) suggest the workability of the proposed model. Overall, findings suggest that when the shear strength of the soil cannot be determined directly, the proposed hybrid XGBoost-SSA model can be utilized to assess this parameter.