<p>The work systematically evaluated oxygen mass transfer efficiency (OTE₂₀) in Venturi flumes using both dimensional and non-dimensional datasets. A range of models, including Bagging (BG) and Boosting (BS) regressors, as well as existing relations, were assessed using tenfold cross-validation. Model performance was evaluated through performance metrics such as the coefficient of correlation (CC), root mean square error (RMSE), mean absolute error (MAE), and graphical analyses. The Random Forest Bagging model (BG_RF) demonstrated the highest predictive accuracy and reliability yielding favourable results for both dimensional datasets (CC: 0.83362, RMSE: 0.07865, MAE: 0.00274) and non-dimensional datasets (CC: 0.81093, RMSE: 0.08288, MAE: 0.00187). The Extreme Gradient Boosting Regressor (BS_XGBR) followed closely, with performance metrics of CC: 0.83160, RMSE: 0.07919, MAE: 0.00408 for dimensional datasets, and CC: 0.79985, RMSE: 0.08534, MAE: 0.00452 for non-dimensional datasets. In comparison, existing relations displayed significantly lower accuracy, with smaller CC values and higher RMSE and MAE values. Uncertainty analysis validated BG_RF's robustness, exhibiting the narrowest uncertainty bands (dimensional: 0.30890; non-dimensional: 0.32565). One-way analysis of variance,&#xa0;established machine learning model accuracy, with generally good model agreement.Sensitivity analyses (one factor at a time, Morris, Sobol) and correlation/Shapley analyses identified discharge per meter width (q) and gauge reading (H_a) as crucial features influencing dimensional variables, while Reynolds number (R_e) and the ratio H_b/H_a were significant in non-dimensional datasets. These factors critically influence OTE₂₀ in Venturi flumes by affecting turbulence, agitation, gas–liquid interface, flow regimes, and oxygen dissolution.</p>