Purpose: To evaluate and compare the capability of ANFIS (Adaptive Neuro-Fuzzy-Inference System), ANN (Artificial Neural Network), and MNLR (Multiple Non-linear Regression) techniques in the estimation and formulation of Discharge Correction Factor (Cd) of modified Parshall flumes as based on linear relations and errors between input and output data. Design/methodology/approach: Acknowledging the necessity of further research in this field, experiments were conducted in the Hydraulics Laboratory of Civil Engineering Department, National Institute of Technology, Kurukshetra, India. The Parshall flume characteristics, associated longitudinal slopes and the discharge passing through the flume were varied. Consequent water depths at specific points in Parshall flumes were noted and the values of Cd were computed. In this manner, a data set of 128 observations was acquired. This was bifurcated arbitrarily into a training dataset consisting of 88 observations and a testing dataset consisting of 40 observations. Models developed using the training dataset were checked on the testing dataset for comparison of the performance of each predictive model. Further, an empirical relationship was formulated establishing Cd as a function of flume characteristics, longitudinal slope, and water depth at specific points using the MNLR technique. Moreover, Cd was estimated using soft computing tools; ANFIS and ANN. Finally, a sensitivity analysis was done to find out the flume variable having the greatest influence on the estimation of Cd. Findings: The predictive accuracy of the ANN-based model was found to be better than the model developed using ANFIS, followed by the model developed using the MNLR technique. Further, sensitivity analysis results indicated that primary depth reading (Ha) as input parameter has the greatest influence on the prediction capability of the developed model. Research limitations/implications: Since the soft computing models are data based learning, hence the prediction capability of these models may dwindle if data is selected beyond the current data range, which is based on the experiments conducted under specific conditions. Further, since the present study has faced time and facility constraints, hence there is still a huge scope of research in this field. Different lateral slopes, combined lateral- longitudinal slopes, and more modified Parshall flume models of larger sizes can be added to increase the versatility of the current research. Practical implications: Cd of modified Parshall flumes can be predicted using the ANN- based prediction model more accurately as compared to other considered techniques. Originality/value: The comparative analysis of prediction models, as well as the formulation of relation, has been conducted in this study. In all the previous works, little to no soft computing techniques have been applied for the analysis of Parshall flumes. Even the regression techniques have been applied only on Parshall flumes of standard sizes. However, this paper includes not only Parshall flume of standard size but also a modified Parshall flume in its pursuit of predicting Cd with the help of ANN and ANFIS based prediction models along with MNLR technique.