<p>Parkinson’s Disease (PD) is a chronic neuro-degenerative disorder that typically targets the neurons in the human brain. In the early stages of PD, speech impairments were discovered in the majority of patients. This study proposes a two-stage ensemble model to diagnose the disease in its initial stage utilizing speech signals. Firstly, an extra tree classifier features selection technique is utilized to discover the most appropriate features from the dataset and the dataset is balanced by the synthetic minority oversampling technique. Finally, the classification problem is addressed by the amalgamation of three base learners, K-nearest neighbor, decision tree, and Naive Bayes, and a new stacked ensemble model is constructed to detect PD. The effectiveness and robustness of the proposed work have been evaluated on accuracy, sensitivity, specificity, and F-1 score. The detection accuracy obtained by the proposed method is 98.31%. The results demonstrate that, in contrast to the traditional feature selection and single classifier technique, the proposed method can effectively decrease the number of features while preserving a higher rate of accuracy of detection. The suggested method surpasses the current cutting-edge studies conducted in this area.</p>