Analysis of Ensemble of Machine Learning Algorithms for Detection of Parkinson's Disease

Parkinson's Disorder(PD) is a neurodegenerative motion disorder. This disorder can be identified by the signs and symptoms which may increase with a small tremor in right or left palms and a feeling of tension. It may be worse thereafter. In overall, six million people in the world are affected by this disease as per the analysis of researchers. The proposed concept is to search for a way to discover Parkinson's sickness at a preliminary level that can assist in supplying the required remedy at the best time and avoid the disorder. Machine Learning algorithms are implemented onto Parkinson's Disease Speech features data set to identify whether the individual is healthy or not. If affected, the severity can also be diagnosed and probably be taken care of appropriately. The proposed work uses various models like Naive Bayes, Random Forest classification, K Nearest Neighbour, XGBoost, Decision Tree, Support Vector Machine. The number of features is also reduced by using Principal component analysis (PCA). The performance of these models are analysed and compared. These models show the accuracy in the range of 70% to 90% using Parkinson's Disease Speech Features dataset. The higher performance models are considered for ensemble and to improve the overall performance metrics. The ensemble of models are used to get the best accuracy of 91% and precise recognition of the disease.