Enhancing Parkinson's disease prediction using meta-heuristic optimized machine learning models.

Abstract

Parkinson's disease is a progressive neurological disorder affecting movement and cognition. Early detection is crucial but challenging with traditional methods. This study applies meta-heuristic optimization to enhance machine learning prediction models. A Parkinson's dataset with demographic, lifestyle, medical, clinical, and cognitive features was analyzed using three feature selection techniques: Whale Optimization Algorithm, Artificial Bee Colony Optimization, and Backward Elimination (BE). Random Forest (RF) models were optimized using Artificial Ant Colony Optimization for hyperparameter tuning. The optimized RF model with BE achieved 93% accuracy and 97% AUC, outperforming K-Nearest Neighbors, Support Vector Machines, Logistic Regression, XGBoost, and Stacked Ensemble models. Optimization reduced tuning time from 133 to 18 minutes. A comparison with traditional approaches and negative controls validated the results, though clinical validation remains essential before deployment. Meta-heuristic optimization significantly improves Parkinson's prediction performance and efficiency.