Comparison and Evalution of Bayesian Neural Network, Gene gramming, Support Vector Machine and Multiple expression proLinear Regression in River Discharge Estimation (Case Study: Sufi Chay Basin)

Prediction of river discharge is importance for reliable planning, design and management of water resources projects. This study investigates the applicability of Bayesian Neural Network (BNN), Gene-Expression Programming (GEP), Support Vector Machine (SVM), and Multi linear Regression (MLR) for prediction of river discharge time series in the Soufichay river, Iran. Daily river discharge time series for period of 1997 to 2010 of Tazehkand hydrometric station from Soufichay river was used. To obtain the best input–output mapping, different input combinations of antecedent daily river discharge were evaluated. The performance of the models were evaluated through the four performance criteria: Correlation Coefficient (CC), Root Mean Square Error (RMSE), the Nash–Sutcliffe efficiency coefficient (N-S) and Bias criteria. A comparison of models indicates that BNN with CC (0.991), RMSE (0.031m3/s), N-S (0.981) and Bias(-0.006) predicted better than GEP, SVM, and MLR models for daily river discharge time series.