Approximation of aeration efficiency at sharp-crested weirs using metaheuristic regression approaches

Abstract This paper explores the ability of multivariate adaptive regression splines, decision trees, Gaussian processes, and multiple non-linear equation approaches to predict aeration efficiency at various weirs discusses their results. In total, 126 experimental observations were collected in laboratory, which 88 arbitrarily selected for model training, rest used validation. Various graphical presentations goodness-of-fit parameters assess performance models. Performance evaluation results, Whisker plot, Taylor's diagram indicated that GP_rbf-based was superior other implemented models predicting with CC (0.9961 0.9973), MAE (0.0079 0.0195), RMSE (0.0122 0.0251), scattering index (0.0594 0.1238), Nash Sutcliffe (0.9923 0.9564) values training validating stages, respectively. The predicted by GP_rbf lie within ±30% error line most it lying at/close agreement. random forest had better predictability than tree implied. sensitivity analysis suggests shape factor drop height as major influencing factors efficiency.