Algorithms based on binary trees can be used for applications involving regression prediction and classification. Decision trees, boosted trees, and bagged trees are among the algorithms examined in this study. The study analyzed the algorithmic structure and developed a procedure for analyzing and predicting a hydraulic parameter these algorithms. The Buckingham Pi theorem was applied to define the objective function, thereby identifying the input data fields required for the machine learning models and determining the experimental data to be collected. The proposed machine learning models were subsequently applied to predict the hydraulic jump characteristics in a trapezoidal channel. The results indicate that machine learning-based modeling is highly suitable for predicting hydraulic parameters. Evaluation using statistical performance indicators showed strong to very strong predictive performance, with the coefficient of determination R² ranging from 0.95 to 1.00. Other statistical indicators, including RMSE, MSE, and MAE, were all very close to their ideal value of zero. The study further showed that the proposed Boosted Tree 7 model (with “Minimum Leaf Size” equal to 1 and “Number of Learners” equal to 800) achieved the best predictive performance. For the training dataset, the model produced an R² value approximately equal to 1, while the other statistical indicators were nearly zero. For the testing dataset, the model also achieved a very strong coefficient of determination, with R² approximately equal to 0.98. The maximum prediction error for hydraulic jump length was 5.2%. These findings demonstrate that machine learning algorithms are highly suitable and effective for predicting hydraulic characteristics in hydraulic engineering applications.