Ensemble Classifier to Enhance Computer Aided Diagnosis of Parkinson Disease

Abstract

Parkinson's disease is a neurodegenerative disorder of the central nerve system which affects movements. Data from 42 persons having early stage of Parkinson's disease was collected with a total number of 5875 voice recordings present in dataset. The different machine learning models were used to predict the motor Unified Parkinsons disease rating score (UPDRS) score from the various voice measures. Then the actual and predicted values for various evaluation parameters (Correlation, R Square, RMSE, Accuracy) are calculated and results are compared. After comparing the results of the various models, the top 5 models are ensembled and results are calculated to give stronger overall prediction. The aim of ensembled model is to calculate the UPDRS from various voice measures with higher accuracy of 99.6%. K-Fold validation approach is used to measure the robustness of ensembled model.