Weirs are hydraulic structures with various applications, such as flow measurement, flow diverting, and/or flow control. Labyrinth weir as a nonlinear weir with a folded crest enhances the spillway system discharge capacity by increasing the crest length for a given channel width. As a remarkable parameter, the discharge coeffcient plays an important role in determining weirs’ passing capacity. So, it is important that the discharge coeffcient be accurately represented to ensure an appropriate and economical design. This research evaluates the prediction accuracy of four soft computing techniques, such as support vector machines (SVM), gene expression programming (GEP), software (QNET), and artifcial intelligence networks (ANN), to predict the discharge coeffcient of labyrinth weirs using three different scenarios. The model’s input parameters included the total water head ratio (HT/P), magnifcation (Lc/W), and cycle wall angle (α). On reviewing the outcomes of the developed models, it appeared that each of them demonstrated satisfactory performance in predicting the discharge coeffcient. However, ANN and SVM reveal more accuracy compared to other techniques. The prediction results reveal that the ANN as a superior model has the following performance metrics: R2 = 0.998, DC coeffcient = 0.996, and RMSE = 0.006. The sensitivity analysis results indicate that HT/P is the most effective factor in determining the labyrinth weir discharge coeffcient in all methods. This suggests that the ratio of HT (total water head) to P (weir height) has a signifcant impact on the overall model performance. Upon comparing the results of this research with those of other researchers, it is evident that the evaluation indicators for all methods employed in this study are relatively superior.
