Intelligent System for Early Detection of Parkinson’s Disease Using ML Techniques

Abstract

 Early detection of Parkinson's disease (PD) is critical for timely treatment and better patient outcomes. Medical professionals may now benefit from well-designed procedures that utilize biological voice metrics to help in the diagnosis of PD, thanks to advanced AI and ML capabilities. With the use of the UC Irvine PD dataset, this study introduces a smart ML approach to early PD detection. The planned approach consists of data preprocessing, feature normalization (StandardScaler), addressing class imbalance (Borderline SMOTE) and subject-level stratified data splitting for unbiased model evaluation. A discriminative pattern is learned from the voice-based features using an XGBoost classifier for accurate classification of healthy individuals or Parkinson's patients. The proposed model was evaluated and found to be as accurate as 98.50%, outperforming other machine learning models. Precision (PRE), recall (REC), F1 score (F1), and confusion matrix analysis were used to further evaluate the robustness and reliability of the suggested framework. The results show that the suggested AI system can help doctors make better decisions by simplifying the diagnostic process and allowing for earlier detection, faster treatment, and better clinical management of PD.