GIS and ML Integrated Techniques for Detection of Landslide-Prone Areas

Landslides are a common natural disaster that can result in substantial casualties. One of the region’s most vulnerable to landslides in India is Western Ghats. The risk casualty is reduced with the use of the landslide susceptibility map (LSM) and prediction of landslide points. The aim of this research is to explore and compare the 4 ML algorithms (K-Nearest Neighbor (KNN), Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR)) for generating the most promising LSM. This study focuses on 12 Landslide conditioning factors (LCFs) such as Elevation, Slope, Lithology, LULC, Distance to Road, Topographic Wetness Index (TWI), Stream Power Index (SPI), Topographic Roughness Index (TRI), Drainage Networks, NDVI, Rainfall, Sediment Transport Index (STI). It has been found that the slope of the ground, the distance to the road, and the annual rainfall all have a significant impact on the occurrence of landslides. The models made using different approaches have been analysed and compared using the following metrics: prediction accuracy, precision, recall, F1-score, and True Skill Statistics (TSS). Overall accuracy for LR, SVM, KNN, RF are 82.30, 84.00, ${8 2. 5 0, 8 7. 2 0}$ respectively. The findings of the study show that there are more landslide-prone locations in Bhor, Velhe, Ambegaon, and Junnar in Pune District, Maharashtra.