Evaluation of neuro-fuzzy GMDH-based particle swarm optimization to predict longitudinal dispersion coefficient in rivers

In the present research, neuro-fuzzy-based group method of data handling (NF-GMDH) has been applied to evaluate the longitudinal dispersion coefficient in rivers. The NF-GMDH model has been improved through particle swarm optimization algorithms (PSO). Effective parameters on the longitudinal dispersion coefficient including flow depth, channel width, cross-sectional average velocity, and bed shear velocity were selected to characterize a correlation between input and output variables. Field and experimental data sets have been collected from different studies. The efficiency of the proposed NF-GMDH-PSO model for both training and testing stages has been investigated. The performance of the NF-GMDH-PSO model were compared with those obtained from the differential evolutionary (DE), model tree (MT), genetic algorithm (GA), artificial neural network (ANN), and traditional empirical equations. Results analysis showed that among the artificial intelligence approach-based equations, DE and GA methods performed better than the other methodologies. The most accurate empirical equations were also introduced. NF-GMDH-PSO network also predicted the longitudinal dispersion coefficient properly and can be considered as an alternative to the aforementioned successful formulas.