Optimizing machine learning and bagging-based hybrid models for landslide susceptibility mapping: a case study in Chenggu County, China.

The study employed four different models, including the bivariate statistical model Certainty Factor (CF), machine learning models such as Functional Tree (FT), Logistic Model Tree (LMT), Alternating Decision Tree (ADT), and Reduced Error Pruning Tree (REPT), hybrid models based on bagging techniques (Bagging-FT, Bagging-ADT, Bagging-REPT, Bagging-LMT), and various machine learning optimization models, to map landslide susceptibility in Chenggu County with the aim of exploring assessment results that offer greater social value for the study area. The field survey combined with remote sensing technology acquired 181 landslides in Chenggu County. The comprehensive research area currently selects 16 landslide conditioning factors and employs the Certainty Factor (CF) method to examine the relationship between landslides and these factors, aiming to elucidate the mechanisms by which these conditioning factors trigger landslides. Grid search methodology is used to conduct hyper-parametric FT, ADT, LMT and REPT and hybrid models of Bagging algorithm for optimization study to obtain the optimal parameter combination of the model and improve the predictive ability of the model. The CF model, machine learning model and optimization model were constructed, and then the Receiver operating Characteristic (ROC) curves and statistical parameters were compared to verify the model accuracy and generalization ability, and the hybrid model prediction accuracy of Bag-LMT after hyperparametric optimization was the best. The results further validate the importance of optimization of machine learning models in geohazard prediction research.