Machine learning approaches for predicting undrained shear strength of soil: A case study with a user-friendly interface

Soil shear strength, a pivotal metric in civil engineering, signifies the soil's capacity to endure shear stress before failing. The accurate determination of this measure is crucial for assessing the stability of structures situated on or embedded within the soil. Traditional methods, while integral, often prove complex and resource intensive. This has paved the way for advanced machine learning (ML) techniques to offer innovative solutions. This study delves into the efficacy of multiple ML algorithms, including Support Vector Regression (SVR), Decision Tree (DT), Random Forest Regression (RFR), Extra Tree Regressor (ETR), Gradient Tree Boosting (GTBR), and Extreme Gradient Boosting (XGBoost) in predicting undrained shear strength of soil. A comprehensive dataset, sourced from prior research, was utilized, with a strategic split of 80% for training and 20% for testing with 5 folds cross validation. Model performance was gauged using statistical metrics such as MAPE, MAE, RMSE, and 𝑅. The findings highlight GTBR as the most proficient predictive model with 𝑅 of 85.52%. Feature importance analysis revealed that variables such as the liquidity index, sample depth, and moisture content percentage played pivotal roles in shaping the model's predictions. This model has been seamlessly integrated into an online user-friendly interface, facilitating ease of access for professionals. The interface ensures a streamlined, precise tool for estimating soil shear strength.