Evaluation of Machine Learning Algorithms (RF and SVM) in Producing FloodSusceptibility Mapping Maroon Watershed

Floods are one of the most important natural hazards that have caused economic and social damage in most areas. Numerous climatic, hydrological, geomorphological and geological factors are involved in the occurrence of floods. Flood analysis, management and control can be done by preparing flood potential maps. The purpose of this study is to map the flood potential of Maroon watershed using random forest and support vector machine machine learning methods. For this purpose, 16 effective parameters in flood occurrence including altitude, slope and aspect, curvature, geological formations, land use, curve number, rainfall, temperature, stream power index )SPI), topographic wetness index )TWI), distance From stream, stream density, road distance, road density and NDVI index were considered. The mentioned parameters were prepared in ArcGIS 10.8, ENVI 5.3 and SAGA GIS 7.2 software environments and then converted to readable format for R software environment in order to implement the support vector machine and random forest models. Finally, RF and SVM models were implemented using SDM packages and evaluated using receiver operating characteristic )ROC). The results showed that RF and SVM models correctly predicted the flooding map of Maroon Basin with 0.997 and 0.947 percent accuracy, respectively