Customer churn is a critical issue in the telecommunications industry, and accurately predicting churn can help companies implement strategies to retain customers. In this study, we propose an advanced stacking model to enhance churn prediction. The proposed model consists of two levels: base models such as Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM, trained using K-fold cross- validation, while the proposed model utilizes a training dataset generated from the outputs of the base models. The advanced stacking model incorporates sophisticated techniques such as cross-validation, SMOTE for handling class imbalance, and custom preprocessing steps to improve accuracy. The evaluation results demonstrate the superior performance of the proposed model, achieving an accuracy of 97.36%, a precision of 97.66%, and an AUC Score of 97.34%, significantly outperforming Ensemble Voting, StackingClassifier, and the base models. These results showcase the enhanced predictive capabilities of the advanced stacking model, highlighting its markedly higher accuracy, precision, and AUC Score compared to the other models.