Reservoir risk modelling using a hybrid approach based on the feature selection technique and ensemble methods

Flash flooding is a type of global devastating hydrometeorological disaster that seriously threatens people’s property and physical safety, as well the normal operation water conservancy facilities, such reservoirs, so an accurate assessment reservoir risk for certain areas necessary. Therefore, purpose this study was to propose novel methodological approach modelling based on feature selection method (FSM) tree-based ensemble methods (Bagging Random Forest [RF]). The results showed that: (1) J48-GA models achieved higher learning predictive capabilities compared conventional without FSM. (2) For classification accuracy, J48-GA-RF (96.4%) outperformed RF (96.0%), J48-GA-Bagging (93.9%) Bagging (93.5%). And highest prediction AUC value (0.995), almost perfect Kappa indexes (0.926) best practicality (30.88%). (3) In particular, indicated all high performance, both in training validation dataset. Additionally, could provide reference managers, hydraulic engineers policy makers implement location-specific flash flood reduction strategies.