predicting streamflow using data-driven model and time series

accurate forecasting of streamflows has been one of the most important issues as it plays a key role in allotment of water resources. river flow simulations to determine the future river flows are important and practical. given the importance of flow in the coming years, in this research three stations: haji qooshan, ghare shoor and tamar in gorganrood cachment were simulated in 2002-2011. to simulate river flow, time series (auto regression) and data driven based on support vector machine (svm) was used for both monthly and weekly. the results showed that both methods in tamar have low precision and haji qooshan station have good precision in monthly simulation. svm increase 0.29 coefficient determination and decreases 0.35 rmse error in ghare shoor station and perform more accurate than time series. both methods simulate weekly discharge in low precision in tamar and ghare shoor. coefficient determination of time series is 0.91 and svm is 0.86 in weekly simulation. ddr statistics show that the svm has greater precision than time series in monthly simulation and equal precision in weekly simulation in haji qooshan station. the results of this study show that the svm method is more accurate than time series in monthly and weekly simulation. the accuracy of both methods is on monthly basis rather than weekly. the accuracy of both methods is greater on monthly rather than weekly.