A comparative Study of Data Mining Techniques for Estimating Suspended Sediment Load in River Flow

Abstract

Abstract-Estimating of Suspended Sediment Load (SSL) in rivers is extremely important for planning and managing of the water resource projects. Support Vector Machine (SVM) and Gene Expression Programming (GEP) techniques are used for estimating SSL in four kilometers along Tigris River, located in upstream Al-Amarah Barrage; Maysan Province; Southern Iraq. Twenty-sections are selected for the purpose of the field measurement of SSL, which include measurement of flow velocity. For applying main object of this research, measured river velocities at these sections are used as the input variables of data mining techniques and the model output is SSL at these river sections. Cross validation method is used to estimate the performance of the models results, a random set of rows is selected to each validation fold after stratifying on the target variable. Three statistical parameters (root mean square error, mean absolute error and coefficient of correlation) are used to evaluate the performance of models. The performances of SVM model are better than GEP model. Data mining techniques specifically (SVM and GEP) are efficient and powerful techniques for modeling suspended sediment load.