Performance Evaluation of GIS-Based Artificial Intelligence Approaches for Landslide Susceptibility Modeling and Spatial Patterns Analysis

摘要: The main purpose of this study was to apply the novel bivariate weights-of-evidence-based SysFor (SF) for landslide susceptibility mapping, and two machine learning techniques, namely naive Bayes (NB) Radial basis function networks (RBFNetwork), as benchmark models. Firstly, by using aerial photos geological field surveys, 263 locations in area were obtained. Next, identified landslides randomly classified according ratio 70/30 construct training data validation models, respectively. Secondly, based on inventory map, combined with geomorphological characteristics area, 14 affecting factors determined. predictive ability selected evaluated LSVM model. Using WoE model, relationship between analyzed positive negative correlation methods. above three hybrid models then used map susceptibility. Thirdly, ROC curve various statistical (SE, 95% CI MAE) verify compare power Compared other Sysfor model had a larger under (AUC) 0.876 (training dataset) 0.783 (validation dataset). Finally, quantitatively comparing values each pixel, differences spatial morphology maps compared, found have limitations effectiveness. obtained are reasonable, generated highest comprehensive performance. results paper can help local governments land use planning, disaster reduction environmental protection.