Predictive Modeling and Early Detection of Parkinson's Disease Using Machine Learning

Abstract

Parkinson's disease is a progressive neurodegenerative disorder that impacts millions of citizens in the USA, mainly targeting the motor system and causing debilitating symptoms such as rigidity, tremors, and bradykinesia. The diagnosis of Parkinson's disease is presently heavily dependent on clinical evaluations and neurological examinations, targeting the detection of motor dysfunction. The principal aim of this study was to use machine learning as a means for early detection and prediction of Parkinson's disease. The dataset utilized for this study was the Parkinson’s Disease Dataset, retrieved from the UCI Machine Learning Repository, which included comprehensive biomedical voice measurements from a cohort of individuals, 23 of whom are diagnosed with Parkinson’s disease and 8 who are healthy controls. This dataset included a set of features extracted from voice recordings. These include parameters like fundamental frequency (pitch), amplitude variation, jitter, shimmer, and several phonation-related measures known to reflect early vocal impairments associated with the disease in question. This research project deployed three credible and proven algorithms, namely, logistic regression, random forest, and the Support Vector Machines. Besides, this study employed a combination of metrics, including accuracy, precision, recall, F1-score, and ROC-AUC, which are more holistic toward the model performance. According to the metric performance results of the three models, several key insights were drawn into the early detection of Parkinson's Disease. Particularly, it was clear that the Random Forest model had superior accuracy and was the most reliable in classifying positive cases and healthy patients, which could be that this model turned out to be most reliable in an early detection setting. In that respect, predictive modeling in Parkinson's Disease is a capability frontier that has the potential to make much difference in clinical decision-making. Supported by Machine Learning algorithms, clinicians can trace the minute patterns in patient data, which are not easily visible through any other diagnostic means. In such cases, early detection of Parkinson's Disorder with vocal biomarkers or motor assessment may afford healthcare professionals opportunities for early interventions that might retard the disease process.