Chapter 15 - Random vector functional link network based on variational mode decomposition for predicting river water turbidity

Abstract

In the present chapter, we propose a new modeling framework for predicting river turbidity using only river discharge as predictor using three models, i.e., random vector functional link neural network (RVFL), generalized regression neural network (GRNN), and the radial basis function neural network (RBFNN). First, the models were applied using only river discharge, i.e., RVFL, GRNN, and RBFNN. Second, the variational mode decomposition (VMD) was used for decomposing the river discharge into several intrinsic mode functions (IMF); thus, hybrid models were developed. At hourly time scale, the RVFL_VMD exhibited the high performances, significantly, superior to the value obtained using the RVFL, i.e., 0.870 and 0.756, respectively. At daily time scale, the GRNN_VMD was the most accurate model having R and NSE of 0.990 and 0.979, respectively. It is argued that the VMD is a robust tool for improving the accuracies of machines learning models used for river water turbidity prediction.