Nonlinear ensemble techniques and reliability to estimate oxygen transfer efficiency of rectangular venturi flumes for free flow

In this study, we investigate assessing the oxygen transfer efficiency of a rectangular venturi flume (VFOTF20), employing machine learning ensemble models. Our investigation revolves around how models like gradient boosting machine (GBM), stacked ensemble (SE), extreme gradient boosting (XRT), M5 (Pruned and Unpruned), and random forest (RF) estimate the oxygen transfer efficiency of a rectangular venturi flume. Key input parameters, including discharge per unit width (q), throat width (W), throat length (F), and water depth (Ha), are scrutinized. Laboratory experiments furnish the datasets essential for our analysis. During the testing phase, these modelling ensembles are rigorously evaluated. The findings unequivocally demonstrate that the GBM model surpasses all other contenders in performance. Nevertheless, it is noteworthy that all proposed ensemble models, alongside conventional models like linear regression (LR) and nonlinear regression (NLR), exhibit commendable performance, albeit with variations. The correlation diagram and sensitivity analysis underscore the pivotal role of discharge per unit width (q) as the most significant variable impacting the VFOTF20. Additionally, we conduct a one-way analysis of variance (ANOVA) and an uncertainty study to enrich our understanding of the subject matter further.