Classification of Parkinson's Disease using Speech Attributes with Parametric and Nonparametric Machine Learning Techniques

Parkinson's disease is a neurological disorder that affects the central nervous system which results in improper functioning of body movements including difficulty in speaking, walking, rigidity, and so on. The aim is to classify Parkinson's disease based on acoustic voice features in dysphonic speech disorder in Parkinson's patients using state of art machine learning algorithms. Some of the parametric and nonparametric machine learning techniques was used on the voice features dataset. This research aims to evaluate the performance of state of art algorithms such as the Naïve Bayes Algorithm, Logistic regression, K-Nearest Neighbors, and Random Forest Algorithm to detect Parkinson's disease which gives higher classification accuracy. The performance is evaluated by pre-processing the data based on speech attributes and fed into the machine learning models for further processing. Various performance metrics like precision, recall, F1 score are computed for all four machine learning techniques and the results show that nonparametric models using Random Forest and K-Nearest Neighbors produce higher classification accuracy of 87.2% and 90.2% compared to parametric models.