Landslide Susceptibility Assessment Using Data Mining Models, A case study: Chehel-Chai Basin

The current study is aimed to map landslide susceptibility in the Chehel-Chai Basin is located in the Golestan Province. To this end, two data mining models namely support vector machine and boosted regression tree were employed due to their robust computational algorithm. Landslide inventories were recorded through several field surveys using global positioning system (GPS), local information and available organizational resources and corresponding map was created in the geographic information system (GIS). Reviewing several worldwide studies, 12 predisposing factors including proximity to fault, proximity to stream, proximity to road, lithological units, soil texture, land use/ cover, slope degree, slope aspect, altitude, plan curvature, profile curvature and topographic wetness index (TWI) were chosen and the corresponding maps were produced in the geographic information system. In order to evaluate models’ results the receiver characteristic curve (ROC) and 30% of landslide inventories were used. Results showed that the SVM model with the AUC value of 0.82 had better performance on landslide susceptibility zonation over the study area and followed by the BRT model with the value of 0.77. Based on the SVM model results, about 45% of the Chehel-Chai Basin has high and very high landslide prone areas.