GIS-based landslide susceptibility modeling using data mining techniques

Introduction: Landslide is one of the most widespread geohazards around world. Therefore, it necessary and meaningful to map regional landslide susceptibility for mitigation. In this research, maps were produced by four models, namely, certainty factors (CF), naive Bayes (NB), J48 decision tree (J48), multilayer perceptron (MLP) models. Methods: first step, 328 landslides identified via historical data, interpretation remote sensing images, field investigation, they divided into two subsets that assigned different uses: 70% subset training 30% validating. Then, twelve conditioning employed, altitude, slope angle, aspect, plan curvature, profile TWI, NDVI, distance rivers, roads, land use, soil, lithology. Later, importance each factor was analyzed average merit (AM) values, relationship between occurrence various evaluated using (CF) approach. next based on effect models quantitatively compared receiver operating characteristic (ROC) curves, area under curve (AUC) non-parametric tests. Results: The results demonstrated all can reasonably assess susceptibility. Of these CF model has best predictive performance (AUC=0.901) validating data (AUC=0.892). Discussion: proposed approach an innovative method may also help other scientists develop in areas could be used geo-environmental problems besides natural hazard assessments.