Enhancing Water Quality Modeling Using Support Vector Regression with Emperor Penguin and Golden Eagle Optimizers

Accurate prediction of the Sodium Adsorption Ratio (SAR) in river systems is essential for effective irrigation management, salinity control, and sustainable water quality practices. This study investigated the predictive capabilities of a standalone Support Vector Regression (SVR) model alongside two hybrid approaches, including SVR optimized by the Emperor Penguin Optimizer (SVR EPO) and SVR optimized by the Golden Eagle Optimizer (SVR GEO), applied to the Qezel Ozan River, Iran. Four input scenarios were developed, ranging from a single input variable (Cl) to a combination of CI, EC, SO۴, and pH. The results revealed that both hybrid models substantially outperformed the traditional SVR model in all scenarios. Notably, in Scenario ۱, SVR GEO achieved the best performance with a root mean square error (RMSE) of ۳.۰۰۷, followed closely by SVR EPO with an RMSE of ۳.۰۹۹, while the original SVR recorded an RMSE of ۷.۰۷۴. These outcomes highlight the superior optimization capacity of the GEO algorithm in improving SAR prediction accuracy. Furthermore, the excellent performance of the simplest scenario emphasizes the dominant role of chloride in influencing SAR values. Overall, the SVR GEO model offers a reliable and accurate approach for forecasting SAR and can serve as a valuable decision-support tool in water resource management and agricultural planning.

کلیدواژه ها:

Machine learning ، Prediction ، Sodium Adsorption Ratio ، Water Quality