AI-HydSu: An advanced hybrid approach using support vector regression and particle swarm optimization for dissolved...

Area of Science:

• Environmental Science
• Data Science

Background:

• Dissolved oxygen (DO) concentration prediction is crucial for the fishing industry.
• Variations in DO are influenced by multiple factors, leading to nonlinear and fuzzy uncertainties.
• Accurate DO prediction remains a significant challenge.

Purpose of the Study:

• To develop a hybrid model (AI-HydSu) for accurate dissolved oxygen concentration prediction.
• To enhance the optimization capabilities for Support Vector Regression (SVR) model parameters.
• To address the nonlinear and fuzzy uncertainties in DO prediction.

Main Methods:

• Data preprocessing using wavelet threshold denoising.
• Employing Particle Swarm Optimization (PSO) with nonlinear adaptive learning factors to optimize SVR parameters (penalty factor, kernel function, error parameter).
• Developing a hybrid AI-HydSu model integrating denoised data and optimized SVR.

Main Results:

• The proposed AI-HydSu model demonstrates high prediction accuracy for dissolved oxygen concentration.
• The optimized PSO algorithm effectively prevents local optima and accelerates the search process.
• The model achieved a fast convergence rate in experimental validation.

Conclusions:

• The AI-HydSu model offers a robust solution for accurate dissolved oxygen prediction.
• Optimized PSO with adaptive learning factors enhances SVR model performance.
• This approach provides a valuable tool for the fishing industry by improving DO monitoring.