Parkinsons disease (PD) is a progressive neurode- genera- tive illness that mostly impacts the motor domain or movement, albeit it can also impact patients’ non-motor domain. The most widely used instrument for assessing the intensity of motor symptoms and monitoring the course of the disease is the Unified Parkinson’s Disease Rating Scale (UPDRS). Since Parkinson’s disease results in motor impairments that affect the muscles of the vocal cords, voice data provides an easily obtainable non-invasive source of data for predicting UPDRS scores. Researchers evaluate the predictive performance of four different machine learning models derived from voice measures of the Unified Parkinson’s Disease Rating Scale (UPDRS). Support Vector Machine (SVM), k-Nearest Neighbors (k-NN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) are among the models that are being studied. According to the data, CNN performs the best among the evaluated models since it has the highest accuracy, precision, recall, and F1 score. In particular, CNN performs very well at this task as highlighted by an accuracy score very close approximating 95.57% and with high precision and recall approximating 95%. The two models K-NN and LSTM, have a performance with balanced scores 93.44 and 92.47 respectively regarding all the metrics, which indicates good accu- racy of the model to understand the intersection of the temporal and frequency-specific features of the voice data. The SVM model obtained the lowest scores in each of the regards, achieving an accuracy of about 74%, demonstrating the lesser strength of this model in identifying appro- priate voice features for predicting UPDRS. This analysis highlights the potential of CNN as a reproducible model to predict UPDRS and open avenues toward accurate, non-invasive, remote monitoring of Parkinson’s disease progression. We will continue optimizing these models and test additional voice features to achieve better predictive performance.