Maryam Asadi; Ali Fathzadeh; Roohollah Taghizadeh Mehrjerdi
Volume 4, Issue 10 , June 2017, , Pages 121-143
Abstract
The main purpose of this study is an inquiry into the functions of daily, monthly, and annual scales of sediment data in their estimations using machine learning models. For this purpose, suspended sediment load data for three temporal, daily, monthly, and annual, scales at Ohio station, located in the ...
Read More
The main purpose of this study is an inquiry into the functions of daily, monthly, and annual scales of sediment data in their estimations using machine learning models. For this purpose, suspended sediment load data for three temporal, daily, monthly, and annual, scales at Ohio station, located in the USA, between the years of 1992 and 2014 were selected. In order to choose the best model, some machine learning base models such as artificial neural networks, error back propagation as well as radial basis function, k-nearest neighbor, M5 decision tree, Gaussian process, support vector machine (SVR), evolutionary support vector machine (ESVM), and linear regression (LR) models were run and evaluated. The results of this study showed that the k-nearest neighbor with RMSE=5.28, the data Gaussian process model with RMSE=8.7, and the Gaussian process model with a RMSE=7.2 were respectively the best models for the daily, monthly, and annual data. The comparison of the models' assessment also suggested that the predicted annual data were more accurate than the monthly and daily data.