Test smells are manifestations of sub-optimal design decisions made during the development of test cases. A considerable body of empirical research has highlighted their detrimental impact on the maintainability and effectiveness of test code. To address this issue, various detection approaches have been introduced, ranging from automated rules/heuristic techniques to machine learning (ML)-based classifiers. Despite these efforts, existing detection techniques often suffer from limited predictive performance and a strong dependent on manually tuned thresholds. In this study, a novel ensemble-based ML model is proposed for test smell prediction, leveraging Stacking Ensemble integrated with k-fold cross-validation to improve accuracy. As the result of the conducted experiments, the proposed model outperforms the state-of-the-art techniques in terms of predictive accuracy, achieving F1-score of 0.8903, ROC of 0.9783, and AUC-PR of 0.9248 for Mystery Guest, while for Resource Optimism, they are 0.8885, 0.9771, and 0.9179, respectively. These findings highlight the superior effectiveness of the proposed approach in the domain of automated test smell detection.