Comparative Analysis of Machine Learning Methods and a Physical Model for Shallow Landslide Risk Modeling

Shallow landslides restrict local sustainable socioeconomic development and threaten human lives property in loess tableland. Therefore, the appropriate creation of risk maps is critical for mitigating shallow landslide disasters. The first task to be done was evaluate vulnerability based on a machine learning model (random forest (RF), support vector (SVM) logistic regression (Log)), physical (SINMAP) tableland area. By comparing differences, best method evaluating selected. nonlinear response relationship between environmental factors quantified frequency ratio. Multicollinearity analysis used identify 10 that were applied ML construct spatial distribution model. SINMAP DEM soil parameters determine stability coefficient study results showed (1) Dongzhiyuan mainly occurred shady slopes with an elevation 1068–1249 m, slope gradient 36°–60° concave shape. stream power transport indexes increased increasing rainfall erosion, making likely. susceptibility changed parabolically change NDVI grassland shrubland. (2) four methods performed similarly predicting sensitivity landslides. high-incidence areas both sides eroded gully slopes. bottom not prone (3) highest area under curve (AUC) values generated from RF training validation datasets 0.92 0.93, respectively, followed by SVM AUC 0.91 0.92, respectively; Log 0.89, 0.69 0.74, respectively. In conclusion, predicted provide scientific basis disaster mitigation Loess Plateau.