Parkinson's Disease Classification using Quantile Transformation and RFE

Abstract

Parkinson's disease (PD) occurs because of the insufficiency of dopamine that manages diverse activities of the human body. Analysts have recognized that voice is a fundamental symptom of PD. The paper mainly focuses on analyzing the characteristic feature selection with RFE on the Quantile transformed data for the regulation of voice analysis of PD systems, along with the use of various classifiers. In this work, we have implemented various machine learning models like the SVM, Random forest, Decision tree, and logistic regression, and accomplished PD classification. The results obtained by using the quantile transform and RFE feature selection, and by tuning the classifiers using hyperparameter tuning with the GridSearchCV method, we have obtained higher accuracy of 90% for SVM and F1-score of 80%, recall of 82%, and precision of 79%. The findings strongly guide the application of the SVM classifier for voice analysis based PD classification with the quantile transform with RFE and GridSearchCV.