TXT-tool 1.081-6.1 A comparative study of the Binary Logistic Regression (BLR) and Artificial Neural Network (ANN) models for GIS-based spatial predicting landslides at a regional scale

Abstract

This teaching tool is to present how to generate the landslide susceptibility maps using binary logistic regression (BLR) and artificial neural network (ANN) methods at a regional scale. The study area is one of most landslide-prone areas in Japan. First, the landslide inventory data from the National Research Institute for Earth Science and Disaster Prevention (NIED) were randomly partitioned into two parts: training and testing data. Then, 10 m DEM data and geology map were analyzed to extract the landslide predisposing factors. Next, the susceptibility maps were produced in a geographic information system (GIS) environment. Then, the receiver operating characteristics (ROC) was used to assess the model accuracy. Validation results show that both of two methods can be obtained with acceptable results. The maps can provide useful information for the future planning of hazard mitigation.