نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Assessing groundwater quality for irrigation purposes in arid and semi-arid regions plays an important role in the sustainable management of water resources. In this study, three hybrid machine learning models, namely MLP-XGBoost, MLP-CatBoost, and MLP-LightGBM, were used to predict the Irrigation Water Quality Index (IWQI) in the Varamin Plain aquifer. The results showed that, among the five parameters constituting the IWQI, sodium (Na⁺), chloride (Cl⁻), electrical conductivity (EC), and sodium adsorption ratio (SAR) are sufficient for predicting the IWQI, and that the index can be predicted with very high accuracy in the Varamin Plain aquifer without considering bicarbonate (HCO₃⁻). All three models performed well, but among the three hybrid models, MLP-XGBoost showed the best performance on the test data (R² = 0.994 and RMSE = 1.158). SHapley Additive exPlanations (SHAP) showed that the most important derived feature in all three hybrid models was the output predicted by the multilayer perceptron (MLP). These findings demonstrate the effectiveness of hybrid stacking ensemble architectures for predicting the Irrigation Water Quality Index, while emphasizing that the trained models are specific to the hydrochemical conditions of the study area and require recalibration before being applied in other regions.
کلیدواژهها English