Implementation of Ensemble Predictive Models for Parkinson’s Disease Detection

A framework based on ensemble classifier models is presented in this study, which may be used to predict the presence of disease and to increase the accuracy of the prediction. Machine learning's ensemble technique pools several different classification algorithms and an extensive number of classifier models to get more precise predictions. In this present study SVM and DT are employed as base estimators, which takes use of well-known ensemble techniques, such as bagging and boosting, to construct ensemble classifiers. The proposed technique is tested on a Parkinson's database taken from machine learning repository and is found to be effective. Classification accuracy (CA), precision, recall, area under curve (AUC), Cohen kappa score (CKS), Mathews correlation coefficient (MCC), and F1-measure obtained from receiver operating characteristics (ROC), as well as the MCC and F1-measure indicate that the ensemble classifier model has enhanced prediction performance with the least possible errors.