Predicting Liquefied Soil Settlement Using Boosting-Based Machine Learning Models

Soil liquefaction-induced settlement is a critical issue in geotechnical engineering due to its potential to cause severe structural damage. Traditional prediction methods often lack accuracy and adaptability when handling complex, nonlinear relationships in soil behavior. In this study, we explore the effectiveness of five boosting-based machine learning models—AdaBoost, Gradient Boosting Machine (GBM), XGBoost, LightGBM, and CatBoost—for predicting post-liquefaction settlement based on geotechnical input parameters. A real-world dataset containing key soil properties and corresponding settlement measurements was used for training and evaluation. The performance of the models was assessed using multiple metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). Among the models, CatBoost demonstrated the highest prediction accuracy with an R² score of 0.9705 on the testing set, outperforming both traditional regression techniques and other ensemble models. The findings confirm the potential of boosting algorithms, particularly CatBoost, in accurately modeling complex soil behavior, offering a valuable tool for engineers in liquefaction risk assessment and mitigation planning.