Elham Rezaei; Babak Shahinejad; Hojatola Yonesi
Abstract
One of the important issues in rivers qualitative discussion is the prediction of amount of Total Dissolved Solids (TDS) in water. In this study, the performance of the intelligent models Support Vector Machines (SVM) with different kernel functions, Gene Expression Programming (GEP) and Bayesian Network ...
Read More
One of the important issues in rivers qualitative discussion is the prediction of amount of Total Dissolved Solids (TDS) in water. In this study, the performance of the intelligent models Support Vector Machines (SVM) with different kernel functions, Gene Expression Programming (GEP) and Bayesian Network (BN) was investigated in the prediction of amount of total dissolved solids (TDS) in Kashkan River. For this purpose, quality parameters obtained monthly from Poldokhtar station in Lorestan Province from year 1991 to 2016, including hydrogen carbonate, chloride, sulfate, magnesium, calcium, sodium, electric conductivity, flow rate and PH these parameters were applied to predict the amount of total dissolved sediments of water in this reach. Correlation Coefficient (CC), Nash-Sutcliff coefficient (NS), Root Mean Square Error (RMSE), and bias were used to evaluate the performance of the models. The results showed that in all three models, the combined structures have acceptable accuracy. Also, based on the evaluation criteria, Support Vector Machines with kernel of the radial base functions (RBF) had the highest accuracy 0.982 and the lowest root mean square error (mg / lit) 0.232, and the lowest bias 0.001 and the Nissan Sutcliff coefficient 0.963 compared to other models.
