Performance Evaluation of ANFIS and M5 Tree Models in Crump Weir Discharge Coefficient Modeling

Crump weir is a kind of short edged weir. This weir with its particular design is used to measure the flow discharge in channels. The experimental dataset used in this research are taken from the others to study the performance of ANFIS and M5 tree models for predicting discharge coefficient (Cd) in crump weir. The 174 experimental data from physical models, considered as 16 scenarios with input parameters consist of Fr, Re, h1/p, Sup and Sdo; and Cd as output using for M5 tree model and ANFIS for modeling and evaluation of the coefficient of crump weir. The obtained results show that both methods give almost near results in scenarios with equal training and testing dataset. The best scenario in ANFIS model achieved for randomized dataset with 70% of training and 30% of testing as scenario 12 with the highest correlation coefficient (R=0.988) and the lowest error (RMSE=0.069). Also the best scenario of M5 tree model for randomized data with 75% of training and 25% of testing as scenario 1 with highest correlation coefficient (R=0.988) and lowest error (RMSE=0.007) was calculated. The results suggest a better performance by the ANFIS approach with this dataset, but M5 model trees, provides simple linear relations for prediction of discharge coefficient for the data ranges used in this study.