Development of Categorical Boosting Regression on Peak Discharge from Earth–Rock Dam Failures Assessment

Accurate peak flow (Qp) estimation at earth-rock dam breaches is vital for flood management. The study collected 162 cases, removed 3 outliers using IQR, leaving 159 records for analysis. The model received 80% of the dataset for training and the remaining 20% as test material. The essential input parameters included the dam type, together with failure mode, dam height, and water height above the breach invert and stored water volume above the breach at failure. The research evaluates Categorical Boosting Regression (CBR) as an ML technique that predicts peak discharge values. The research team applied Starfish Optimization (StFO) and Tyrannosaurus Rex Optimization (TyRO) metaheuristic optimizers to select hyperparameters because these choices determine how well the model will perform. Results show that CBRStFO outperforms CBRTyRO, achieving higher R² values in both training (0.979) and testing (0.967), with prediction improvements of about 4%, demonstrating superior accuracy and reliability.