Data-Driven Approach to Parkinson's Disease Detection Using SMOTE and Classifier

In this study, we present a comprehensive approach to improving the detection of Parkinson's disease using machine learning techniques. The research addresses the challenge of data imbalance, prevalent in medical datasets, by incorporating the Synthetic Minority Over-sampling Technique (SMOTE) to enhance model performance. Data was obtained from a publicly available dataset and preprocessed to improve its quality for analysis. Multiple machine learning classifiers, including Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest, Gradient Boosting, and XGBoost, were employed to assess the effectiveness of SMOTE in improving model accuracy and classification metrics. The results demonstrated that models trained with SMOTE consistently outperformed those without SMOTE regarding accuracy, precision, recall, and F1 score. Notably, when SMOTE was applied, XGBoost achieved the highest performance with an accuracy of 70.37%. This study highlights the significance of data balancing strategies in improving the predictive capabilities of machine learning models for Parkinson's detection, with potential implications for developing practical clinical diagnostic tools.