Automated Improved Detection of Parkinson’s Disease using Ensemble Modeling

A person suffering from Parkinson's disease (PD) mostly has speech impairments. Speech impairment in PD refers to the disorder of speech due to motor and non-motor deficits. Many articles have been reported in developing PD detection model by using standard speech datasets. But still there is further scope of improvement in the detection potentiality of PD. In this paper, this issue has been addressed and a novel ensemble method by combining the advantages of two popular classifiers such as Support Vector Machine (SVM) and Random Forest (RF) has been proposed. The contribution of the classifiers in the final prediction is determined by assigning weights which are optimized by Particle Swarm Optimization (PSO) algorithm. Further the performance of the proposed model is evaluated using a popular and standard UCI Oxford Parkinson's Disease Detection Dataset and it is demonstrated that the proposed ensemble models achieves 93.7% and 100% accuracy by using 10-fold cross validation scheme and 30% testing set respectively. Thus the suggested ensemble approach outperforms the existing PD prediction models.