NeuroSignal Precision: A Hierarchical Approach for Enhanced Insights in Parkinson's Disease Classification

Parkinson's disease (PD) is a progressive neurological disorder that affects movement, posture, handwriting, and speech. Parkinson's disease is challenging to diagnose early due to subtle symptoms that often go unnoticed, necessitating reliable and accurate classification models to aid clinical decision-making. This research introduces a comprehensive benchmarking of nine unified models, and a unique contribution of this research is the adaptation of the Tabular Transformer model for structured medical data, achieving an unprecedented accuracy of 99.49%, setting a new benchmark for Parkinson's disease classification. The proposed approach provides an advanced, adaptable framework that supports clinicians in making early, accurate diagnoses, ultimately improving patient care. In contrast to previous studies that predominantly emphasize traditional models, this research employs attention-based deep learning to capture complex feature interactions, achieving substantially higher accuracy. The study evaluates nine models: SVM, Decision Tree, Random Forest, AdaBoost, Gradient Boosting, XGBoost, KNN, CNN, and Tabular Transformer, achieving improved accuracy across all models compared to previous studies, marking a notable advancement in Parkinson's disease classification performance. The Transformer's attention mechanism captures intricate data patterns, providing clear advantages over traditional approaches and improving diagnostic precision for early-stage Parkinson's detection. Data preprocessing included the Synthetic Minority Over-sampling Technique for class balancing and feature standardization, with each model, from SVM and Decision Trees to CNN and XGBoost, optimized through Optuna for optimal performance. This research offers the medical field a versatile, high-accuracy framework that aids clinicians in timely and reliable PD diagnosis, potentially improving patient outcomes and advancing Parkinson's diagnostic tools for future clinical use.