Ensemble methods have gained significant popularity in the field of machine learning due to their ability to improve predictive performance by combining multiple models. One ensemble technique that has shown promising results is ensemble stacking, which involves training a meta-model on predictions from multiple base models. However, this method is highly underutilized in applied research. This research focuses on investigating and optimizing ensemble stacking while also incorporating tailored sampling techniques for imbalanced datasets. The study explores a variety of data balancing strategies, including undersampling, oversampling, and hybrid approaches, to mitigate class imbalance. Two ensemble meta-learners were considered, evaluating their ability to capture the underlying class distributions and mitigate bias while maintaining overall model performance. The obtained results indicate that the ensemble model with RF as the meta-model achieved the most significant improvement in accuracy of 94%. The research findings from this study will contribute to the development of optimized ensemble stacking techniques for addressing imbalanced classification challenges, enabling improved decision-making and performance in real-world applications.