Assessing ANFIS accuracy in estimation of suspended sediments

Abstract

Capabilities offered by an adaptive neuro-fuzzy inference system (ANFIS) in the estimation of daily sediment loads at four stations in the USA, are explored in the paper. For this purpose, models with various input combinations of data sets were constructed to enable identification of the best possible structure. The results show that the best ANFIS model exhibits better performance compared to the SRC model, in terms of the RMSE, MBE and R2 values. The results also indicate that the ANFIS model can be applied to facilitate modelling of nonlinear dynamics of complex systems.