Landslide susceptibility mapping using support vector machine for Meghalaya, India

The northeastern region of India is known for its heavy rainfall-induced landslides, yet studies on the area are limited. We have picked Meghalaya, one of the eight northeastern states, as our study location. We aimed to create a comprehensive landslide susceptibility map for Meghalaya using cuttingedge machine-learning techniques. For this purpose, we used the support vector machine, a supervised classification method trained with 1005 existing landslide data, an equal number of non-landslide points generated using a custom toolbox, and 19 landslide causative factors to generate the susceptibility maps. The result was two maps, one with an AUC value of 0.90 and the other with feature selection and an AUC value of 0.91, displaying excellent prediction accuracy. Our landslide susceptibility map can be a valuable disaster planning and management tool for Meghalaya.