A Wavelet Support Vector Machine Combination Model for Daily Suspended Sediment Forecasting

In this study, wavelet support vector machine (WSWM) model is proposed for daily suspended sediment (SS) prediction. The WSVM model is achieved through combination of two methods; discrete wavelet analysis and support vector machine (SVM). The developed model was compared with single SVM. Daily discharge (Q) and SS data from YadkinRiver at Yadkin College, NC station in the USA were used. In order to evaluate the model, the root mean square error (RMSE), mean absolute error(MAE) and coefficient of determination (R-2) were used. Results demonstrated that WSVM with RMSE =3294.6 ton/day, MAE=795.22 ton/day and R-2 =0.838 were more desired than the other model with RMSE =6719.7 ton/day, ton/day and R-2=0.327. Comparisons of these models revealed that, MAE and error standard deviation for WSVM model were about 40% and 50% less than SVM model in test period.