A Comparative Analysis of Certainty Factor-Based Machine Learning Methods for Collapse and Landslide Susceptibility Mapping in Wenchuan County, China

After the “5·12” Wenchuan earthquake in 2008, collapses and landslides have occurred continuously, resulting accumulation of a large quantity loose sediment on slopes or gullies, providing rich material source reserves for occurrence debris flow flash flood disasters. Therefore, it is great significance to build collapse landslide susceptibility evaluation model County local disaster prevention mitigation. Taking as research object according data 1081 historical points, well natural environment, this paper first selects six categories environmental factors (13 total) including topography (slope, aspect, curvature, terrain relief, TWI), geological structure (lithology, soil type, distance fault), meteorology hydrology (rainfall, river), seismic impact (PGA), ecological (NDVI), human activity (land use). It then builds three single models (LR, SVM, RF) CF-based hybrid (CF-LR, CF-SVM, CF-RF), makes comparative analysis accuracy reliability models, thereby obtaining optimal area. Finally, study discusses contribution prediction model. The results show that (1) areas prone extremely high predicted by CF-LR, RF CF-RF) an area 730.595 km2, 377.521 361.772 372.979 318.631 306.51 respectively, frequency ratio precision 0.916, 0.938, 0.955, 0.956, 0.972, 0.984, respectively; (2) ranking comprehensive index based confusion matrix CF-RF>RF>CF-SVM>CF-LR>SVM>LR AUC value CF-RF>RF>CF-SVM>CF-LR>SVM>LR. To certain extent, coupling can improve more over models. CF-RF ranks highest all indexes, with POA 257.046 0.946; (3) rainfall, river are most important factors, accounting 24.216%, 22.309%, 11.41%, respectively. necessary strengthen monitoring mountains rock masses close rivers case rainstorms county other similar post-earthquake landslides.