Parkinson’s Disease Detection using ML in the Context of Overfitting Issue

This study leverages machine learning (ML) models on biomedical voice data for Parkinson’s disease detection, evaluating Decision Tree (DT), Random Forest (RF), XGBoost, Logistic Regression (LR), and Support Vector Machine (SVM) for accuracy, precision, recall, and F1 score. The DT model, for instance, achieves 100% across all metrics in training but drops to 77% accuracy and 74% recall in testing, indicating significant overfitting. Similarly, RF’s accuracy declines from 100% to 79%, with 84% recall on testing data. XGBoost achieves 87% accuracy and recall, while LR and SVM exhibit minimal overfitting, with SVM achieving 88% accuracy and 96% recall. Overall, SVM demonstrates the most reliable performance for PD detection.