The prediction of longitudinal dispersion coefficient in natural streams using LS-SVM and ANFIS optimized by Harris...

Area of Science:

• Environmental Science
• Water Resource Management
• Computational Fluid Dynamics

Background:

• Accurate longitudinal dispersion coefficient (Kx) calculation is crucial for river pollution modeling.
• Existing experimental equations often lack reliability due to uncertainties and missing data.
• Machine learning (ML) models offer a promising alternative for Kx prediction but are underutilized.

Purpose of the Study:

• To predict the longitudinal dispersion coefficient (Kx) in rivers using ML models.
• To compare the performance of ML models against traditional experimental methods.
• To evaluate the efficacy of ANFIS optimized by Harris Hawk Optimization (ANFIS-HHO).

Main Methods:

• Utilized Least Square-Support Vector Machine (LS-SVM), Adaptive Neuro-Fuzzy Inference System (ANFIS), and ANFIS-HHO.
• Employed various input variable combinations, including flow depth (H), flow velocity (U), and shear velocity (u*).
• Compared ML model predictions against experimental methods using metrics like RMSE, MAPE, and R-squared.

Main Results:

• ML models demonstrated superior performance over experimental equations in predicting Kx.
• ANFIS-HHO achieved the best performance, with RMSE of 17.0, MAPE of 0.22, and R-squared of 0.97.
• The Harris Hawk Optimization (HHO) algorithm enhanced ANFIS prediction accuracy.

Conclusions:

• ML models, especially ANFIS-HHO, provide highly precise Kx predictions, outperforming experimental methods.
• Experimental equations tend to overestimate Kx values, while ML models offer greater accuracy.
• The ANFIS-HHO approach is a robust tool for river pollution modeling and other environmental applications.