Prediction for Parkinson's Disease using Machine Learning through Voice

Parkinson's Disease (PD) is traditionally diagnosed through clinical assessments, primarily focusing on motor symptoms, which can result in delayed detection. These systems often fail to account for non-motor symptoms, such as subtle vocal changes, that may precede more apparent signs of the disease. To address this gap, we propose a machine learning approach utilizing voice analysis for the early prediction of PD. This study applies Support Vector Machines (SVMs) to analyze vocal features like jitter, shimmer, and Harmonics-to-Noise Ratio (HNR) from a dataset of 195 voice samples collected from 31 participants, 23 of whom are diagnosed with PD. Using SVM with a linear kernel, we achieved an accuracy of 87% and a precision of 91% in classifying individuals with PD based on vocal characteristics. This non-invasive technique highlights the potential of vocal feature analysis as an early diagnostic tool for PD, allowing for earlier interventions. Our study demonstrates the effectiveness of SVM in high-dimensional data spaces and its robustness against overfitting, offering a promising enhancement to current diagnostic methods for Parkinson's Disease.