Dimensionality Reduction Method for Prediction of Parkinson Disease using Speech Data
International Journal for Research in Applied Science and Engineering Technology
Neurological diseases are risky if the symptoms are not detected. It is important to detect the symptoms accurately at an early stage. Parkinson's disease (PD) is a significant nervous system disorder affecting ten million people worldwide. Parkinson disease's can be identified by three characteristics: speech, memory and movement disorder. Speech and memory data is used to predict the presence of the disease. Movement disorders are a less reliable resource for detection. Collection and
... lection and processing of speech data are less intensive. Since memory disorder symptoms are detected at a later stage, the challenge increases. Generally, classification and regression techniques are implemented for the identification of PD patients. The proposed work focuses on using Principal component analysis on a combination of speech and memory information. The technique is used for reducing the dimension of the features. Traditional classification algorithms have been applied and compared. The accuracy of the dimensionality reduction techniques is improved to 97% by using the Random Forest approach.