Parkinson’s disease is a neurodegenerative disorder that affects movement. Diagnosing Parkinson’s disease has traditionally involved clinical assessments by neurologists, and this practice still persists today to a significant extent. However, clinical assessments can be prone to subjectivity. In this study, a comprehensive predictive modeling approach was undertaken, employing nine dis¬tinct machine learning algorithms and six different model evaluation metrics to identify the best per¬forming algorithms. The findings reveal that, using only 12 vocal characteristics, KNeighborsClassfier (KNC), MLPClassifier (MLP), and XGBClassifier (XGBC) achieved the highest score of 0.87. This score is generally considered very good, indicating that the model is robust and possesses strong predictive power. This study marks a crucial initial step in leveraging machine learning techniques for more effective and potentially more accurate diagnosis of Parkinson’s disease based on patients’ vocal characteristics.
Research Article
Open Access