Prediction of flow discharge in compound sections, Comparison of empricial and data driven methods

Document Type : Original Article

Authors

Msc student of water structures of shahid bahonar university of kerman,kerman,iran

Abstract

Rivers discharge prediction in condition of situation appearance of compound channel is one of the important parameters in the flood and rivers engineering.in this investigation compound channel discharge has been predicted by using data drivens such as artificial neural network and support vector machines (calissification, regression) and also empricial method of channel divided by vertical division.For this purpose collected 150 expremental data frome 6 scientific source. this data divided to four categories twenty five percent by using k fold cross validation.in each modeling in data driven approaches, model testing performed by one group of data.After averageing of the same results in each data driven approach, the results showed the superiority of the SVR method .this method  has maximom of  , NS(equal to respectively 0.92875,0.898075) and has minimum of RMSE,MAPE (equal to respectively 0.08435( ),23.89%) and also has percentage maximom of relative superiority of RMSE than artificial neural network was performed more successful than other methode in the compund channel discharge prediction.

Keywords


Rivers discharge prediction in condition of situation appearance of compound channel is one of the important parameters in the flood and rivers engineering.in this investigation compound channel discharge has been predicted by using data drivens such as artificial neural network and support vector machines (calissification, regression) and also empricial method of channel divided by vertical division.For this purpose collected 150 expremental data frome 6 scientific source. this data divided to four categories twenty five percent by using k fold cross validation.in each modeling in data driven approaches, model testing performed by one group of data.After averageing of the same results in each data driven approach, the results showed the superiority of the SVR method .this method  has maximom of  , NS(equal to respectively 0.92875,0.898075) and has minimum of RMSE,MAPE (equal to respectively 0.08435( ),23.89%) and also has percentage maximom of relative superiority of RMSE than artificial neural network was performed more successful than other methode in the compund channel discharge prediction.