Modeling of river flows with support vector machines and fuzzy logic methods

Accurate determination of the amount of flow in rivers is important for efficient use of water resources, planning the construction of water structures and preventing flood disasters. However, accurate flow estimation is about a good understanding of the hydrological and meteorological characteristics of the river basin. Estimation of flow in rivers is very important in terms of using water resources for their intended purpose, designing and planning water structures and taking precautions against natural disasters. Today, different artificial intelligence methods are applied for flow prediction in rivers. In this study, the performance of Support Vector Machines (SVM), Fuzzy Logic (Sugeno and Mamdani) and Multiple Linear Regression (MLR) methods were compared and the flow rate in rivers was estimated. The Stilwater river in the USA Worcester Country was chosen as the study area. 1095 daily precipitation, flow and temperature data were used to generate the input data in fuzzy logic, support vector machines and multiple linear regression models. Obtained results and actual results were compared.