Abstract
Abstract
Hydrologic practices and other hydrological applications can be conducted successfully only when the stream flow behavior in a river watershed is estimated accurately. In-depth use of several machine learning techniques has been made to comprehend this hydrological phenomenon. In cases of in-depth research on the comparison of machine learning algorithms, the literature is still lacking. This study compares the performance of Support Vector Regression (SVR) and Artificial Neural Network (ANN) in rainfall-runoff modeling of the Awash Belo Watershed. The technique of optimal model input selection for the Machine learning method has been assessed using Auto Correlation and Cross-Correlation Functions. The optimal model input for this research was rainfall and discharge with their lag one and two. Four criteria have been chosen to assess the consistency between the recorded and predicted flow rates: the Root-Mean-Square Error, the Coefficient of Determination, Nash Sutcliff, and the Mean absolute error. The optimized parameters for these models were selected using the GridSearchCV optimization technique with 10 cross-validations. The daily runoff values computed using SVR and ANN models, and their corresponding daily discharges of 5 years during the testing periods (2001− 2005) were evaluated at R2, NSE, RMSE, and MAE with values 0.95, 0.95, 3.12, and 1.28 for ANN and 0.95, 0.96, 3, and 1.27 for SVR respectively. The two models showed comparable performance. Therefore, both model performs the same and can be applied to the study area to estimate flow rates for further investigation.
Publisher
Research Square Platform LLC
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献