A comparative study of landslide susceptibility maps using logistic regression, frequency ratio, decision tree, weights of evidence and artificial neural network

摘要

For the purpose of comparing susceptibility mapping methods in Mizunami City, Japan, the landslide inventory was partitioned into three groups as various training and test datasets to identify the most appropriate method for creating a landslide susceptibility map. A total of fifteen landslide susceptibility maps were produced using frequency ratio, logistic regression, decision tree, weights of evidence and artificial neural network models, and the results were assessed using existing test landside points and areas under the relative operative characteristic curve (AUC). The validation results indicated that the logistic regression model could provide the highest AUC value (0.865), and a relatively high percentage of landslide points fell in the high and very high landslide susceptibility classes in this study. Furthermore, the paper also suggested that the model performances would be increased if appropriate landslide points were used for the calculation.