Daily Flow Modeling With Random Forest and K-Nearest Neighbor Methods

Water is an indispensable natural resource for life. Therefore, protection and control of water resources are of great importance. Since river flow estimation and modeling are very important in cases such as the management of water resources, irrigation, it is included in the literature as an issue that needs constant research and development. A large number of techniques are being used for estimation and modeling; thus, the estimation results are gradually improving with the development of the studies carried out, the comparison of techniques, and the determination and removal of the shortcomings. In this study, Random Forest and K-Nearest Neighbors nonlinear regression models, which are two of the machine learning methods, were used to evaluating the estimation results, to find the better estimation method, and to determine the advantages and disadvantages of these methods. In addition, Random Search and Grid Search methods were used to make the hyperparameter selection and comparison for the Random Forest model. In this study, in which daily flow data of 1981-2011 of the two stations in the Euphrates were used, and, when compared to other models, it was observed that better results were obtained when Random Search was applied to determine the hyperparameters of the Random Forest model.