Document Type : Research Paper


1 PhD Candidate, Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares University, Noor, Iran

2 Associate Professor, Mountain and Arid Regions Reclamation Department, Faculty of Natural Resources, University of Tehran, Karaj, Iran

3 Assistant Professor, Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran

4 PhD Candidate, Watershed Management Engineering, Department of Watershed Management, University of Kashan, Kashan, Iran


This study aimed to examine the influence of pre-processing input variables by Gamma Test on performance of Support Vector Machine in order to predict the suspended sediment amount of Doiraj River, located in Ilam Province from 1994-2004. The flow discharge and rainfall were considered as the input variables and sediment discharge as the output model. Also, the duration of the model training period was determined through GT. Thereafter, in order to carry out the influence of pre-processing input variables on performance of model, the suspended sediment was predicted using SVM model while no pre-processing has been done on its input variables and the results were compared to each other. Results show the performance of the GT-SVM model in the test phase with minimum RMSE was equal to 0.96 (ton/day) and the maximum coefficient of R2 was equal to 0.98 between the predicted and actual values, was better than SVM model.