Parkinson's disease detection using ensemble techniques and genetic algorithm

Abstract

Parkinson's disease (PD) is a neurological disorder which progress by time. People suffering from PD experience shortage of Dopamine which is a chemical present in brain nerve cells. The symptoms of PD are tremor, rigidity, and slowness of movements and people with PD experience more severity by time progress. Therefore, the automation in early detection of PD is an important issue. In the literature, different classification methods have been proposed. Also, due to the high dimension of extracted features of voice, many feature selection algorithms have been developed. In this paper, we aim to propose a method for early detection of PD from voice recordings. The Genetic algorithm is used to select the optimal set of features which can reduce feature vector dimension from 22 to 6 features. We have achieved 96.55% and 98.28% detection rate by employing AdaBoost and Bagging algorithms for classification process, respectively.