Parkinson’s Disease (PD) is a progressive neurological disorder that impairs both motor and non-motor functions, significantly reducing the quality of life. The early symptoms of PD are difficult to detect, and clinical evaluations often remain subjective. Researchers have developed an automatic PD detection system using voice-based biomarkers as key input features. We collected voice data from 31 subjects, including 23 PD patients, extracting 22 acoustic features such as fundamental frequency (MDVP-Fo), jitter and shimmer measures, and nonlinear complexity metrics (RPDE, DFA, and PPE). The combination of SMOTE and CatBoost with Optuna hyperparameter optimization effectively addresses class imbalance issues in the predictive model. The developed model achieves a high precision of 98.31% in identifying PD patients, outperforming traditional methods. A real-time PD screening solution is integrated into a Flask-based web interface, providing a noninvasive and cost-effective assessment tool for early Parkinson’s disease detection. Our findings confirm that voice analysis is a promising diagnostic tool, with potential improvements when combined with additional data types, including gait patterns and electrical brain signals