Data analysis and machine learning techniques for predicting characteristic movement patterns of Parkinson's disease

Parkinson’s disease (PD) is the second most common neurodegenerative disorder globally, with cases rising exponentially. Despite ongoing efforts, it remains uncured, and accurate diagnosis is still a challenge. Research indicates many diagnoses are incorrect, with only slight improvements in accuracy over the years. To address this, researchers have developed devices that collect inertial data from patients, using sensors to capture movement and identify PD motor symptoms. This study analyzed data from Machado et al. [1], involving PD patients and neurologically healthy individuals, to evaluate a machine learning model for estimating the likelihood of PD-related motor symptoms. A Binary Logistic Regression model was applied, achieving an average accuracy of approximately 90% on out-of-sample data. The model’s performance suggests it could significantly enhance the diagnostic process for PD.