Implementation of Support Vector Machine (SVM) Method in Parkinson's Disease Classification Based on Acoustic Features

Abstract

Parkinson’s disease is a progressive neurodegenerative disorder that significantly impacts quality of life and necessitates accurate early detection. Acoustic analysis of voice features offers a non-invasive and promising approach for classifying this condition. This study aims to evaluate the performance of the Support Vector Machine (SVM) algorithm in classifying Parkinson’s disease based on 22 voice-related features extracted from a public dataset comprising 195 samples. The methodology includes data preprocessing (standardization and class weighting), model training using GridSearchCV, and evaluation based on standard classification metrics and diagnostic curves. The SVM model with an RBF kernel achieved an accuracy of 94.87%, precision of 96.55%, recall of 96.55%, F1-score of 96.55%, and a ROC-AUC score of 0.9828. The results indicate that SVM can effectively handle class imbalance and outliers without the need for complex techniques such as SMOTE or external feature selection. It is concluded that SVM is an effective method for early detection of Parkinson’s disease based on voice data. Future research should focus on testing the model on larger and more diverse datasets and enhancing model interpretability for clinical use.