Study on stability prediction of collapsible cave slopes based on data augmentation and ensemble learning

Aiming at the slope instability and engineering safety issues caused by loess collapsible caves, this study constructed an intelligent slope stability prediction model integrating data augmentation and ensemble learning. To address the limitations of existing prediction models, such as scarce data samples and class imbalance, an innovative data augmentation strategy was proposed: Key features including rainfall, slope gradient, and collapsible cave parameters were selected to build a dataset, and high-quality samples were synthesized by combining the SMOTE algorithm and CTGAN model. This effectively alleviated data imbalance and expanded the coverage of the parameter space. Based on the Stacking ensemble strategy, the advantages of four base models—Random Forest, Gradient Boosting, Support Vector Machine (SVM), and Logistic Regression—were integrated. The Bayesian optimization algorithm was introduced for adaptive hyperparameter tuning. The results show that the optimized ensemble model achieved an accuracy of 90.0%, an F1-score of 0.95, and an AUC value of 0.92 on the test set. Additionally, the model’s prediction results are highly consistent with the calculation results of geotechnical software.