Predict Total Sediment Load Using Standalone and Ensemble Machine Learning Models

Sediment load includes bed and suspended loads. Bed load is sediment on a river’s bottom, while a suspended load is sediment floating in water currents. The nonlinear and multidimensional variables affecting total sediment load make predictions difficult. Ensemble machine learning (ML) models (Additive Regression (AR) and Random Subspace (RSS)), as well as standalone ML models (Locally weighted learning (LWL)), produce a significantly better prediction of the total sediment load when nine input variables from three different properties (sediment, geometry, and dynamic) are combined. Several performance indicators have been used to compare the performance of the models. The input combination IC_7 gave the best result of all the proposed ML models. The hybrid models outperformed the standalone models. The AR-LWL model had the highest prediction accuracy (coefficient of determination ( \(R^2\) ) = 0.91), followed by RSS-LWL, LWL, and the empirical equation in terms of accuracy.