|
چکیده
|
To address the challenges of predicting groundwater quality, we propose an interpretable machine learning approach. We employ advanced algorithms, including XGBoost, Random Forest, GradientBoost, and CatBoost regressors, to develop predictive models for groundwater quality. The SHapley Additive exPlanations (SHAP) method provides insights into water quality parameters’ contributions to the irrigation water quality index (IWQI). Performance metrics like RMSE, MAE, and Rgauge model accuracy, while feature engineering techniques, such as recursive feature elimination with cross-validation (RFECV) and permutation importance (PI), optimize the models’ performance and feature selection. This study, conducted in M’sila state, a semi-arid region heavily reliant on groundwater for agriculture, aims to support sustainable irrigation management. The results will contribute to data-driven strategies for optimizing irrigation practices, ensuring food security, and responsible water resource management.
|