Bagging-based machine learning algorithms for landslide susceptibility modeling

Landslide hazards have attracted increasing public attention over the past decades due to a series of catastrophic consequences landslide occurrence. Thus, mitigation and prevention been topical issues. Thereinto, numerous research achievements on susceptibility assessment springing up in recent years. In this paper, four benchmark models including best-first decision tree (BFTree), functional tree, support vector machine classification regression (CART) were integrated with bagging strategy. Then, these bagging-based applied map regional Jiange County, Sichuan Province, China. Fifteen conditioning factors employed establishing models, respectively, slope aspect, angle, elevation, plan curvature, profile TWI, SPI, STI, lithology, soil, land use, NDVI, distance rivers, roads lineaments. Then utilize correlation attribute evaluation method weigh contribution each factor. Finally, comprehensive performance various corresponding was evaluated systematically compared applying receiver operating characteristic curve area under (AUC) values. Results demonstrated that ensemble significantly outperformed their validation dataset. Among them Bag-CART model has highest AUC value 0.874; however, CART is only 0.766, which reflected satisfying predictive capacity some degree. The obtained study reference values for landslides resource planning County.