<![CDATA[Preeclampsia is a pregnancy complication that poses significant risks to both mother and fetus. Early prediction of preeclampsia risk is crucial to improve maternal healthcare outcomes. This study aims to develop a predictive model for preeclampsia risk using ensemble learning approaches and to enhance model interpretability through Explainable Artificial Intelligence (XAI). The dataset consists of 332 pregnant women who received antenatal care, with 330 complete clinical records after data cleaning. Two ensemble learning algorithms, Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), were implemented and evaluated using Receiver Operating Characteristic (ROC) curves, Area Under the Curve (AUC), and additional classification metrics. The best-performing model was further analyzed using SHapley Additive exPlanations (SHAP) to assess feature contributions at both global and individual levels. The results indicate that XGBoost outperformed Random Forest with an AUC of 0.81 compared to 0.72 after applying class weighting and 5-fold cross-validation. XGBoost also demonstrated more balanced performance with an accuracy of 0.83, recall of 0.85, and specificity of 0.60. In contrast, Random Forest achieved an accuracy of 0.91 and specificity of 0.98 but failed to detect positive cases, with a recall of 0.00, indicating bias toward the majority class. SHAP analysis reveals that height, weight, age at menarche, and the number of antenatal care (ANC) visits significantly influence prediction, while hypertension consistently contributes to increased risk. This study demonstrates that integrating ensemble learning with XAI improves both predictive performance and model transparency for preeclampsia risk assessment.]]>
