K-nearest neighbors and gradient boosting machines are the two data-driven models that are evaluated in this paper for their accuracy and usefulness in predicting the discharge that is going through rectangular planform weirs. A total of 288 experiments were carried out, which resulted in the production of a dataset that included the following ranges: the ratio of crest length to channel width (L/B) from 1.000 to 1.397, the height of the weir crest (P) from 0.08 m to 0.12 m, the head over the weir crest (H) from 0.0312 m to 0.1038 m, and the discharge rate (Q) from 0.0022 m3/s to 0.0129 m3/s. The accuracy, computational efficiency, resilience, interpretability, and scalability of KNN and GBM were evaluated in order to determine their respective levels of performance effectiveness. The results showed that GBM performed better than KNN, achieving reduced mean squared error, root-mean-squared error, and mean absolute error, in addition to achieving a higher R-squared (R2) value. It is possible to ascribe the improved performance of GBM to its capacity to properly handle complicated and nonlinear interactions, as well as its resistance to overfitting. This publication demonstrates that advanced data-driven models, in particular GBM, are effective in precisely estimating the discharge of rectangular planform weirs. As a result, it provides useful insights that may be used to hydraulic engineering applications.