Optimal flood susceptibility model based on performance comparisons of LR, EGB, and RF algorithms

Wadi El-Matulla, located in the eastern desert of Egypt, is most important water basin. The Qift–Qusayr highway (west–east direction) and Cairo–Aswan (north–south pass through watershed. Many urban areas (villages industrial areas) agricultural lands are at outlet these basins. In addition, basin has promising potential for future economic development as it within Golden Triangle (governmental megaproject). current study investigates flood hazard modeling its impact on area. To determine optimal susceptibility mapping algorithm, performance comparisons three techniques were conducted: logistic regression (LR), extreme gradient boosting (EGB), random forest (RF). Remote sensing, topographic, geologic, meteorological data used with help field visits to provide spatial inventory database required by models. reliability predictions proposed models evaluated using five statistical indices: receiver operating characteristic–area under curve, overall accuracy (OAC), kappa index, root mean square error (RMSE), absolute (MAE). showed that values ROC (93, 86 80%), OAC (88, 82 76%), index (0.85, 0.75 0.51), RMSE (0.34, 0.42 0.49) MAE (0.12, 0.18 0.24) RF, EGB, LR, respectively. Based AUC values, RF EGB excellent very good prediction susceptibility. Our results show algorithm mapping, followed LR. Consequently, predictive power model quite map was classified into classes, namely low (51.7%), (23.7%), moderate (16.2%), high (7.1%), (1.3%). Ultimately, verified sentinel-1 imagery real floods 2016 2021, provides agreement. could be useful decision makers planners protect existing facilities plan projects non-flood-prone areas. Accordingly, suitable need distributed mainly