Ensemble-Based Susceptibility Modeling with Predictive Symmetry Optimization: A Case Study from Mount Tai, China

A novel ensemble-based susceptibility modeling approach with predictive symmetry optimization achieves superior accuracy (AUC = 0.964) for geological hazard prediction at Mount Tai, China, outperforming individual machine learning models (max AUC = 0.927). This method addresses small-sample constraints by dynamically weighting models (MLP and XGB) based on output symmetry, reducing bias while maintaining robustness across six algorithms (LR, MLP, RF, GBDT, XGB, SVM). Key drivers of geohazards—elevation, slope, topographic wetness index, and road adjacency—were identified, with high-susceptibility zones concentrated in southeastern terrain and road-intensive areas. The approach combines grid search for hyperparameter tuning and SHAP values for factor contribution quantification, offering a scalable solution for small-sample risk assessment.

Zhao, Z., Chen, B., Liu, P., Duan, X., Ji, Z., Feng, C., Tan, X., Zhang, Y., & Cui, F. (2025). Ensemble-Based Susceptibility Modeling with Predictive Symmetry Optimization: A Case Study from Mount Tai, China. Symmetry, 17(8), 1353. https://doi.org/10.3390/sym17081353