New Machine Learning Ensemble for Flood Susceptibility Estimation

Floods are among the most severe natural hazard phenomena that affect people around world. Due to this fact, identification of zones highly susceptible floods became a very important activity in researcher’s work. In context, present research work aimed propose following 3 novel ensembles estimate flood susceptibility Putna river basin from Romania: UltraBoost-Weights Evidence (U-WOE), Stochastic Gradient Descending-Weights (SGD-WOE) and Cost Sensitive Forest-Weights (CSForest-WOE). regard, sample 132 locations 14 predictors was used as input datasets aforementioned models. The modeling procedure performed through ten-fold cross-validation method revealed SGD-WOE ensemble model achieved highest performance terms ROC Curve-AUC (0.953) also Accuracy (0.94). slope distance importance genesis, while aspect, TPI, hydrological soil groups, plan curvature have lowest influence occurrence. area with high represents between 21% 24% Romania.