This study aimed to evaluate the vulnerability of groundwater to nitrate ( \(\:{\text{N}\text{O}}_{3}^{-}\) ) contamination in the SaissBasin, based on the drinking water standard of 50 mg/L. An innovative approach was adopted, combining the Analytic Hierarchy Process (AHP), artificial intelligence (AI) algorithms including Convolutional Neural Networks (CNN), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LGBM), along with the SHapley Additive exPlanations (SHAP) tool, in order to improve the accuracy and interpretability of vulnerability mapping. Nine environmental and hydrogeological factors were considered, and a dataset of 200 nitrate concentration measurements was used, with 80% allocated for training and 20% for testing. Model performance was evaluated using several statistical indicators: the area under the ROC curve (Receiver Operating Characteristic, AUC), the Kappa coefficient, the root meansquare error (RMSE), and the F1-score. The AHP-CNN model achieved the highest performance (AUC = 0.83), followed by AHP-XGB (0.73) and AHP–LGBM (0.71). SHAP analysis identified Sand, recharge and groundwater depth as the most influential factors. This study aims to identify the most vulnerable areas, as well as the influential factors contributing to nitrate ( \(\:{\text{N}\text{O}}_{3}^{-}\) ) pollution in the SaissBasin, in order to support effective management and protection of groundwater resources.
