Abstract:

The number of computer-aided systems and related studies to provide diagnostic support to medical experts have increased in recent years. Early detection and accuracy of the system play an essential role in early diagnosis of Parkinson’s disease (PD). In this paper, feed-forward backpropogation artificial neural networks (ANNs) with many scenarios are used to predict the Parkinson’s disease through 195 voice recordings obtained from the dataset which consists of voice measurements of 32 people which of 24 have the Parkinson's disease. In the experiment, in addition to ANN it is tried to simulate support vector machine (SVM) model and train it with various train percentages for binary classification to be able to make comparison between ANN and SVM. The highest average accuracy is obtained as 96.88 from the ANN classification algorithm while the highest average algorithm is 87.90% from the SVM classifier when train percentage is 0.9. The highest average ANN test accuracy is 84.33% and achieved with ANN (7-18-1).

Keywords: