Cardiovascular disease is one of the leading causes of death worldwide, making early risk detection essential to support clinical decision-making. Although machine learning models have been widely applied to cardiovascular disease prediction, challenges remain in achieving consistent performance improvements through ensemble learning and providing interpretable prediction results. This study aims to develop a cardiovascular disease risk prediction model by integrating ensemble learning and Explainable Artificial Intelligence (XAI) using the Cardiovascular Disease Dataset from Kaggle, consisting of 70,000 patient records. The research process includes Exploratory Data Analysis, data preprocessing, and model development using Random Forest, CatBoost, and LightGBM combined using a stacking approach with Logistic Regression as the meta-learner. Model interpretability was analyzed using SHAP and LIME, while performance was evaluated using accuracy, precision, recall, F1-score, and AUC-ROC. The results show that LightGBM achieved the highest accuracy of 74.13%, recall of 70.48%, and F1-score of 72.95%, while Random Forest achieved the highest precision of 76.21%. The stacking model achieved the highest AUC-ROC of 80.92%, although it did not consistently outperform individual models across all evaluation metrics. The interpretability analysis identified systolic blood pressure (ap_hi), age, cholesterol, diastolic blood pressure (ap_lo), and body mass index (BMI) as the most influential features in the predictions. This study demonstrates the integration of stacking ensemble learning and XAI using SHAP and LIME, resulting in a model with competitive predictive performance and improved interpretability for cardiovascular disease risk assessment.