Detection of Parkinson´s disease from vocal features using random subspace classifier ensemble

Parkinson´s disease (PD) is a neurological disorder which is diagnosed through clinical examinations and observations rated on Unified Parkinson´s disease Rating Scale (UPDRS). However, in the earlier stages of the disease, this approach might be inconclusive and result in misdiagnosis. Therefore, expert systems are needed to increase the detection accuracy of PD. In this paper, a random subspace based classifier ensemble with k-nearest neighbor (k-NN) as the base classifier was investigated for detection of PD. It was found that the random subspace k-NN classifier ensemble can outperform the single k-NN for a PD recognition problem.