Comparative Study of Classification Algorithms for Early Identification of Parkinson’s Disease Based on Baseline Speech Features

Abstract

- In recent years, there is need for early identification of Parkinson’s disease (PD) based on measuring the features that causes disorders in elderly people. Around 80% of Parkinson’s patients show signs of speech-based disorders in the early stages of the disorder. In this paper, early prediction of Parkinson’s disease based on machine learning is compared between different classification algorithms. The proposed comparative study composed of feature extraction, preprocessing, feature selection and three different classification processes. Baseline features and Iterative Feature selection methods were well thought-out for feature selection process. We compare the performance of classification algorithms used for early prediction of Parkinson’s patients with speech disorders. Naïve Bayes, Multilayer Perceptron, Random Forest and J48 Classification algorithms were used for the categorization of Parkinson's patients in the experimental study. Random Forest and Naïve Bayes classification shows better performance from other two classifiers. 94.1176 % accuracy was obtained from the PD dataset with the smaller number of speech features.