Empirical Analysis of Machine Learning Models on Parkinson’s Speech Dataset

Parkinson’s disease (PD) is a chronic and progressive neurodegenerative disorder that worsens over time. Diagnosing PD primarily relies on clinical assessments, which can be costly, time-consuming, and invasive. These evaluations may also be subjective and vulnerable to inaccuracy. Dysarthria, a common condition characterised by delayed and distorted speech, frequently coexists with PD. This opens up the possibility of using speech features for diagnostic reasons. This research paper explores different machine learning models trained on numerical data of changes in speech patterns due to Dysarthria. These models are based on classifiers such as Artificial Neural Networks (ANN), Multi-Layer Perceptron (MLP), Random Forests, and Decision Trees. Additionally, we compare the performance of a newly introduced HyperTab classifier with the existing models. Our findings demonstrate the significant potential of machine learning in diagnosing PD based on speech analysis. This progress holds the promise of creating a cost-effective tool to expedite disease detection. Furthermore, this research is of utmost importance in offering essential support to regions with limited access to specialized medical facilities.