<![CDATA[Parkinson's disease is a neurodegenerative disorder that affects the nervous system, physiological functions, and brain behavior. Diagnosing this disease remains challenging due to the variability of early symptoms that are difficult to identify. Machine learning has emerged as an effective tool for early detection of Parkinson's disease. Previous studies have implemented various machine learning approaches using different data types; however, research utilizing speech features for early detection remains limited. Furthermore, prior work has predominantly relied on single machine learning models, and no study has yet applied ensemble learning through a stacking technique specifically for detecting Parkinson's disease using speech features. To address this gap, this study proposes a stacking ensemble model that combines three base classifiers, namely K-Nearest Neighbors, Random Forest, and Support Vector Machine, with Logistic Regression serving as the meta-classifier. The proposed model is optimized through hyperparameter tuning employing four strategies, namely Grid Search, Random Search, Bayesian Optimization, and Genetic Algorithm, which are applied to adjust the hyperparameters of the constructed model. Experimental results demonstrate that the optimized stacking model using the best hyperparameters obtained through hyperparameter tuning achieves a superior accuracy of 0.8716 compared with those of three individual classifiers and the stacking model with default hyperparameters. These findings confirm the effectiveness of hyperparameter tuning in improving the proposed stacking ensemble model for detecting Parkinson's disease through speech features.]]>
