Machine Learning-Based Prediction of Tunnel Crown Settlement: A Comparative Study of SVR, XGBoost, Random Forest, and Decision Tree Models

Accurate prediction of tunnel crown settlement is critical for ensuring the safety and stability of underground construction projects. This study systematically evaluates the performance of four machine learning models—Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost), Random Forest (RF), and Decision Tree (DT)—in predicting tunnel crown settlement under complex geological conditions. Data preprocessing techniques, including anomaly detection, linear interpolation, and Savitzky-Golay smoothing, were employed to enhance data quality and ensure reliable model training. The models were comprehensively assessed using multiple evaluation metrics, including MAE, MSE, RMSE, MAPE, MBE, and R2. Results indicate that SVR outperforms other models in terms of prediction accuracy, stability, and generalization capability, making it the optimal choice for tunnel settlement prediction. XGBoost serves as a reliable alternative, while RF and DT exhibit limitations due to overfitting issues. This study highlights the importance of selecting appropriate machine learning models for small-sample datasets and provides valuable insights for engineering applications in tunnel settlement prediction.