<p>This study investigated shallow groundwater in the eastern mountainous region of Tangshan City and developed an integrated evaluation framework combining the Entropy-Weighted Water Quality Index (EWQI) with a Stacking model. Based on 316 water samples and ten hydrochemical indicators (including NO₃⁻, As, and Mn), the EWQI method was initially applied to assess groundwater quality. Subsequently, a Stacking framework was constructed, employing Random Forest, eXtreme Gradient Boosting, Gradient Boosting Decision Tree, and Extremely Randomized Trees as base learners, with Logistic Regression employed as the meta-learner. The EWQI results indicated that groundwater quality was generally good, with only 1.90% of samples exceeding the Class III standard, mainly in densely populated areas such as the Qian’an Basin. The Stacking model achieved an accuracy of 0.947 and a Kappa coefficient of 0.890 on the test set, exhibiting excellent classification performance. Permutation feature importance analysis identified Mn as the most influential indicator, consistent with the findings of hydrochemical analysis. Overall, this integrated system effectively addresses sample imbalance, accurately identifies key factors influencing groundwater quality, and provides a robust scientific basis for regional groundwater management.</p>