Research on SVR Water Quality Prediction Model Based on Improved Sparrow Search Algorithm.

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Multiparameter water quality trend prediction is crucial for effective water environment management and regulation.
• Existing methods face challenges with population diversity and local optima in optimization algorithms.
• Support Vector Regression (SVR) is a powerful tool for regression tasks, but its performance depends on optimal parameter selection.

Purpose of the Study:

• To propose a novel water quality prediction model with enhanced prediction performance.
• To address the limitations of the standard Sparrow Search Algorithm (SSA) in terms of population diversity and local optima.
• To optimize Support Vector Regression (SVR) model parameters using an improved optimization algorithm for superior accuracy and generalization.

Main Methods:

• Development of an Improved Sparrow Search Algorithm (ISSA) by incorporating Skew-Tent mapping for initial population diversity and an adaptive elimination mechanism to escape local optima.
• Integration of ISSA with Support Vector Regression (SVR) to optimize the penalty factor (C) and kernel function parameter (g).
• Comparative performance evaluation of the ISSA-SVR model against BP neural network, SVR, and SSA-SVR using actual breeding-water quality data.

Main Results:

• The ISSA-SVR model achieved a prediction accuracy of 99.2%, significantly outperforming other benchmark models.
• Mean Square Deviation (MSE) was reduced by 79.37% compared to SVR and 75% compared to SSA-SVR.
• The coefficient of determination (R²) reached 0.98, indicating a substantial improvement in model fit and predictive power.

Conclusions:

• The proposed ISSA-SVR model demonstrates superior performance in multiparameter water quality trend prediction.
• ISSA effectively enhances SVR model accuracy and generalization by optimizing key parameters.
• The ISSA-SVR model holds significant engineering application value for water body management and regulation.