Landslide is the movement of materials on the slope (containing natural rocks, soil, artificial accumu-lations or a mixture of them) that are moved downward by the force of gravity. Therefore, the preparation of landslide susceptibility maps is very vigorous so that mitigate damaging effects. The aim of the research is to spatially landslide susceptibility model via three methods of random forest (RF), maximum entropy (ME) and support vector machine (SVM) algorithms, and compare the efficiency of these models in landslide susceptibility mapping in Karganeh Wa-tershed, Lorestan Province, Iran. Landslide inventory map was primed via aerial photographs interpretation and general field surveys. In this research, 16 vital landslide causal factors were reflected to search their spatial connotation to landslide, established happening native geomorphological and human effects. Then, landslide susceptibility maps were built via tree models in geographic information system (GIS). The ROC curve via AUC index was usage to evaluate too compare landslide susceptibility models. Results presented that SVM model provided considerably higher prediction accuracy of landslide susceptibility map in the Karganeh Watershed by ROC equal to 0.913. The subsequent landslide susceptibility maps be able to be appropriate in fitting watershed management practices in this watershed.
Key Words: Landslide susceptibility, Forest, Entropy, Vector Machine, Karganeh Watershed, Iran