OPTIMIZING THE PERFORMANCE OF XGBOOST ALGORITHM IN THE PREDICTION OF PARKINSON'S DISEASE

Parkinson’s Disease (PD) is a progressive neurodegenerative disorder for which early and accurate diagnosis remains a significant clinical challenge. Traditional diagnostic methods rely primarily on subjective clinical evaluations, which are prone to errors especially in the early stages when symptoms overlap with other conditions. These limitations often lead to delayed or inaccurate diagnoses, ultimately compromising patient care and quality of life. Machine learning (ML) models, particularly those analyzing clinical data such as speech patterns, have shown promise in improving diagnostic accuracy. Among them, eXtreme Gradient Boosting (XGBoost) has demonstrated strong performance in medical classification tasks. However, the effectiveness of XGBoost is heavily dependent on proper hyperparameter tuning, as improper settings can result in underfitting or overfitting, thereby reducing model reliability and generalization. This study addresses the need for more precise and robust prediction models for Parkinson’s Disease by implementing XGBoost in combination with hyperparameter optimization using the Grid Search technique. The results show that the optimized XGBoost model significantly outperforms the baseline model (XGBoost without tuning), achieving an improved accuracy of 89% compared to the original 84.6%. This improvement highlights the critical role of hyperparameter tuning and feature selection in building effective predictive models. By leveraging clinical features derived from speech analysis and systematically fine-tuning model parameters, the proposed method enhances the accuracy, reliability, and interpretability of Parkinson’s Disease prediction, offering a valuable tool for early and more accurate diagnosis.