MACHINE LEARNING APPROACH FOR PREDICTING PARKINSON’S DISEASE AT EARLY STAGES

Parkinson’s Disease (PD) is a progressive neurological disorder that often causes speech impairments such as dysphonia, making voice analysis a useful non-invasive diagnostic tool. This study proposes a hybrid machine learning framework using a Voting Classifier that combines K-Nearest Neighbors (KNN), Random Forest, and XGBoost to improve detection accuracy and robustness. The model applies feature normalization and dimensionality reduction to enhance performance. By aggregating predictions from multiple classifiers, the system achieves more reliable and interpretable results, supporting early and effective Parkinson’s disease detection. Furthermore, the proposed system addresses common challenges such as class imbalance and feature variability by leveraging ensemble learning techniques. The integration of multiple models ensures better generalization and stability across different datasets. This approach not only improves prediction performance but also provides a scalable and practical solution for real-world clinical applications, assisting healthcare professionals in timely diagnosis and decision-making.